[ATMOSPHERE-315] Add support of ceph dashboard in grafana (#1709)
This is an automated cherry-pick of #1688
/assign mnaser
diff --git a/roles/kube_prometheus_stack/files/dashboards/ceph-cluster-advanced.json b/roles/kube_prometheus_stack/files/dashboards/ceph-cluster-advanced.json
new file mode 100644
index 0000000..ff31ebf
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/ceph-cluster-advanced.json
@@ -0,0 +1,3813 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "heatmap",
+ "name": "Heatmap",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "singlestat",
+ "name": "Singlestat",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Ceph cluster overview",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CLUSTER STATE",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "0": {
+ "text": "HEALTHY"
+ },
+ "1": {
+ "text": "WARNING"
+ },
+ "2": {
+ "text": "ERROR"
+ }
+ },
+ "type": "value"
+ },
+ {
+ "id": 1,
+ "options": {
+ "match": null,
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "#9ac48a"
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 1
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 2
+ }
+ ]
+ },
+ "unit": "none"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 1
+ },
+ "id": 3,
+ "interval": "1m",
+ "links": [ ],
+ "options": {
+ "colorMode": "value",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_health_status{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Ceph health status",
+ "transparent": true,
+ "type": "stat"
+ },
+ {
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "match": null,
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "max": 1,
+ "min": 0,
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(245, 54, 54, 0.9)"
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 0.10000000000000001
+ },
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": 0.29999999999999999
+ }
+ ]
+ },
+ "unit": "percentunit"
+ }
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 3,
+ "x": 3,
+ "y": 1
+ },
+ "id": 4,
+ "interval": "1m",
+ "links": [ ],
+ "maxDataPoints": 100,
+ "options": {
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "showThresholdLabels": false,
+ "showThresholdMarkers": true
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_cluster_total_bytes{cluster=~\"$cluster\", }-ceph_cluster_total_used_bytes{cluster=~\"$cluster\", })/ceph_cluster_total_bytes{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Available Capacity",
+ "transparent": false,
+ "type": "gauge"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 2,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "match": null,
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)"
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 0.025000000000000001
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 1
+ }
+ ]
+ },
+ "unit": "decbytes"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 6,
+ "y": 1
+ },
+ "id": 5,
+ "interval": "1m",
+ "links": [ ],
+ "options": {
+ "colorMode": "none",
+ "graphMode": "area",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_cluster_total_bytes{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Cluster Capacity",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 1,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "match": null,
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ }
+ ]
+ },
+ "unit": "Bps"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 9,
+ "y": 1
+ },
+ "id": 6,
+ "links": [ ],
+ "options": {
+ "colorMode": "none",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_w_in_bytes{cluster=~\"$cluster\", }[5m]))",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "Write Throughput",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 1,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "match": null,
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "#d44a3a"
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 0
+ },
+ {
+ "color": "#9ac48a",
+ "value": 0
+ }
+ ]
+ },
+ "unit": "Bps"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 12,
+ "y": 1
+ },
+ "id": 7,
+ "links": [ ],
+ "options": {
+ "colorMode": "none",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_r_out_bytes{cluster=~\"$cluster\", }[5m]))",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "Read Throughput",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgb(255, 0, 0)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 0,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "none"
+ }
+ },
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 15,
+ "y": 1
+ },
+ "id": 8,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [ ],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "All",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_osd_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "All",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "In",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_osd_in{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "In",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Out",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_in{cluster=~\"$cluster\", } == bool 0)",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "Out",
+ "refId": "C",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Up",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_up{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "Up",
+ "refId": "D",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Down",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_up{cluster=~\"$cluster\", } == bool 0)",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "Down",
+ "refId": "E",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1
+ }
+ ],
+ "title": "OSDs",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgba(245, 54, 54, 0.9)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 1,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "none"
+ }
+ },
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 6,
+ "w": 3,
+ "x": 21,
+ "y": 1
+ },
+ "id": 9,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [ ],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "Active",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mgr_status{cluster=~\"$cluster\", } == 1) or vector(0)",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "Active",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Standby",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mgr_status{cluster=~\"$cluster\", } == 0) or vector(0)",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "Standby",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ }
+ ],
+ "title": "MGRs",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgba(245, 54, 54, 0.9)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 1,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 1
+ }
+ ]
+ },
+ "unit": "none"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Critical"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "fixedColor": "red",
+ "mode": "fixed"
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Warning"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "fixedColor": "#987d24",
+ "mode": "fixed"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 4
+ },
+ "id": 10,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [ ],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "Active",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ALERTS{alertstate=\"firing\",alertname=~\"^Ceph.+\", severity=\"critical\", cluster=~\"$cluster\", }) OR vector(0)",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "Critical",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Standby",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ALERTS{alertstate=\"firing\",alertname=~\"^Ceph.+\", severity=\"warning\", cluster=~\"$cluster\", }) OR vector(0)",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "Warning",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ }
+ ],
+ "title": "Firing Alerts",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 0.025000000000000001
+ },
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": 0.10000000000000001
+ }
+ ]
+ },
+ "unit": "decbytes"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 6,
+ "y": 4
+ },
+ "id": 11,
+ "links": [ ],
+ "maxDataPoints": 100,
+ "options": {
+ "colorMode": "none",
+ "graphMode": "area",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_cluster_total_used_bytes{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "Used Capacity",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ }
+ ]
+ },
+ "unit": "ops"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 9,
+ "y": 4
+ },
+ "id": 12,
+ "links": [ ],
+ "maxDataPoints": 100,
+ "options": {
+ "colorMode": "none",
+ "graphMode": "area",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_w{cluster=~\"$cluster\", }[1m]))",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "Write IOPS",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colors": null,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [
+ {
+ "id": 0,
+ "options": {
+ "result": {
+ "text": "N/A"
+ }
+ },
+ "type": "special"
+ }
+ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "#d44a3a",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 0
+ },
+ {
+ "color": "#9ac48a",
+ "value": 0
+ }
+ ]
+ },
+ "unit": "ops"
+ }
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 12,
+ "y": 4
+ },
+ "id": 13,
+ "links": [ ],
+ "maxDataPoints": 100,
+ "options": {
+ "colorMode": "none",
+ "graphMode": "area",
+ "justifyMode": "auto",
+ "orientation": "horizontal",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_r{cluster=~\"$cluster\", }[1m]))",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "Read IOPS",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgba(245, 54, 54, 0.9)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 1,
+ "datasource": "$datasource",
+ "description": "",
+ "displayName": "",
+ "fieldConfig": {
+ "defaults": {
+ "decimals": 0,
+ "links": [ ],
+ "mappings": [ ],
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "none"
+ }
+ },
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 15,
+ "y": 4
+ },
+ "id": 14,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [ ],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ },
+ "textMode": "auto"
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "In Quorum",
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_mon_quorum_status{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "In Quorum",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Text Only"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Total",
+ "crit": 1,
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mon_quorum_status{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Total",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Text Only",
+ "warn": 2
+ },
+ {
+ "aggregation": "Last",
+ "alias": "MONs out of Quorum",
+ "crit": 1.6000000000000001,
+ "datasource": "$datasource",
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Annotation",
+ "displayValueWithAlias": "Never",
+ "expr": "count(ceph_mon_quorum_status{cluster=~\"$cluster\", }) - sum(ceph_mon_quorum_status{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "MONs out of Quorum",
+ "range": true,
+ "refId": "C",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1.1000000000000001
+ }
+ ],
+ "title": "Monitors",
+ "transparent": false,
+ "type": "stat"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 7
+ },
+ "id": 15,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "CLUSTER STATS",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "datasource": {
+ "type": "datasource",
+ "uid": "grafana"
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 0,
+ "y": 8
+ },
+ "id": 16,
+ "limit": 10,
+ "onlyAlertsOnDashboard": true,
+ "options": {
+ "alertInstanceLabelFilter": "{alertname=~\"^Ceph.+\", cluster=~\"$cluster\", }",
+ "alertName": "",
+ "dashboardAlerts": false,
+ "groupBy": [ ],
+ "groupMode": "default",
+ "maxItems": 20,
+ "sortOrder": 1,
+ "stateFilter": {
+ "error": true,
+ "firing": true,
+ "noData": false,
+ "normal": false,
+ "pending": true
+ },
+ "viewMode": "list"
+ },
+ "show": "current",
+ "sortOrder": 1,
+ "stateFilter": [ ],
+ "title": "Alerts",
+ "type": "alertlist"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 40,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 0,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "decimals": 2,
+ "thresholds": {
+ "mode": "percentage",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "#c0921f",
+ "value": 75
+ },
+ {
+ "color": "#E02F44",
+ "value": 85
+ }
+ ]
+ },
+ "unit": "bytes"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Total Capacity"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "fixedColor": "red",
+ "mode": "fixed"
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Used"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "fixedColor": "green",
+ "mode": "fixed"
+ }
+ },
+ {
+ "id": "custom.thresholdsStyle",
+ "value": {
+ "mode": "dashed"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 8,
+ "y": 8
+ },
+ "id": 17,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "last"
+ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true,
+ "sortBy": "Last",
+ "sortDesc": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "desc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_cluster_total_bytes{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Total Capacity",
+ "range": true,
+ "refId": "A",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_cluster_total_used_bytes{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Used",
+ "range": true,
+ "refId": "B",
+ "step": 300
+ }
+ ],
+ "title": "Capacity",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 85
+ }
+ ]
+ },
+ "unit": "decbytes"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 16,
+ "y": 8
+ },
+ "id": 18,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "mean",
+ "lastNotNull",
+ "max",
+ "min"
+ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "desc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_w_in_bytes{cluster=~\"$cluster\", }[5m]))",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Write",
+ "range": true,
+ "refId": "A",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_r_out_bytes{cluster=~\"$cluster\", }[5m]))",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Read",
+ "range": true,
+ "refId": "B",
+ "step": 300
+ }
+ ],
+ "title": "Cluster Throughput",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "decbytes"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 0,
+ "y": 16
+ },
+ "id": 19,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "mean",
+ "lastNotNull",
+ "max",
+ "min"
+ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "desc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_w{cluster=~\"$cluster\", }[1m]))",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Write",
+ "range": true,
+ "refId": "A",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_op_r{cluster=~\"$cluster\", }[1m]))",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Read",
+ "range": true,
+ "refId": "B",
+ "step": 300
+ }
+ ],
+ "title": "IOPS",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "bytes"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 8,
+ "y": 16
+ },
+ "id": 20,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "right",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "desc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_pool_bytes_used{cluster=~\"$cluster\", }) *on (pool_id) group_left(name)(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Pool Used Bytes",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ }
+ ]
+ },
+ "unit": "bytes"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "rbd Stored"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "fixedColor": "transparent",
+ "mode": "fixed"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 16,
+ "y": 16
+ },
+ "id": 21,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "table",
+ "placement": "right",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "desc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_pool_stored_raw{cluster=~\"$cluster\", }) *on (pool_id) group_left(name)(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "range": true,
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Pool Used RAW Bytes",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "none"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 0,
+ "y": 24
+ },
+ "id": 22,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_pool_quota_objects{cluster=~\"$cluster\", }) *on (pool_id) group_left(name)(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Pool Objects Quota",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "none"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "bytes"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 8,
+ "y": 24
+ },
+ "id": 23,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_pool_quota_bytes{cluster=~\"$cluster\", }) *on (pool_id) group_left(name)(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Pool Quota Bytes",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": false,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 16,
+ "y": 24
+ },
+ "id": 24,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "right",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "(ceph_pool_objects{cluster=~\"$cluster\", }) * on (pool_id) group_left(name)(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A"
+ }
+ ],
+ "title": "Objects Per Pool",
+ "type": "timeseries"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 31
+ },
+ "id": 25,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "OBJECTS",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 2,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byRegexp",
+ "options": "/^Total.*$/"
+ },
+ "properties": [
+ {
+ "id": "custom.stacking",
+ "value": {
+ "group": false,
+ "mode": "normal"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 12,
+ "w": 6,
+ "x": 0,
+ "y": 32
+ },
+ "id": 26,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "asc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pool_objects{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Total",
+ "range": true,
+ "refId": "A",
+ "step": 200
+ }
+ ],
+ "title": "OSD Type Count",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 2,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byRegexp",
+ "options": "/^Total.*$/"
+ },
+ "properties": [
+ {
+ "id": "custom.stacking",
+ "value": {
+ "group": false,
+ "mode": "normal"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 12,
+ "w": 8,
+ "x": 6,
+ "y": 32
+ },
+ "id": 27,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "displayMode": "table",
+ "placement": "right",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "asc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_active{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Active",
+ "range": true,
+ "refId": "A"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_clean{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Clean",
+ "range": true,
+ "refId": "B"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_peering{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Peering",
+ "range": true,
+ "refId": "C"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_degraded{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Degraded",
+ "range": true,
+ "refId": "D",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_stale{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Stale",
+ "range": true,
+ "refId": "E",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_unclean_pgs{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Unclean",
+ "range": true,
+ "refId": "F",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_undersized{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Undersized",
+ "range": true,
+ "refId": "G",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_incomplete{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Incomplete",
+ "range": true,
+ "refId": "H"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_forced_backfill{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Forced Backfill",
+ "range": true,
+ "refId": "I"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_forced_recovery{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Forced Recovery",
+ "range": true,
+ "refId": "J"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_creating{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Creating",
+ "range": true,
+ "refId": "K"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_wait_backfill{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Wait Backfill",
+ "range": true,
+ "refId": "L"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_deep{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Deep",
+ "range": true,
+ "refId": "M"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_scrubbing{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Scrubbing",
+ "range": true,
+ "refId": "N"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_recovering{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Recovering",
+ "range": true,
+ "refId": "O"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_repair{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Repair",
+ "range": true,
+ "refId": "P"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_down{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Down",
+ "range": true,
+ "refId": "Q"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_peered{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Peered",
+ "range": true,
+ "refId": "R"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_backfill{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Backfill",
+ "range": true,
+ "refId": "S"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_remapped{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Remapped",
+ "range": true,
+ "refId": "T"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_backfill_toofull{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Backfill Toofull",
+ "range": true,
+ "refId": "U"
+ }
+ ],
+ "title": "PGs State",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 2,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "normal"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byRegexp",
+ "options": "/^Total.*$/"
+ },
+ "properties": [
+ {
+ "id": "custom.stacking",
+ "value": {
+ "group": false,
+ "mode": "normal"
+ }
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 10,
+ "x": 14,
+ "y": 32
+ },
+ "id": 28,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "mean",
+ "lastNotNull"
+ ],
+ "displayMode": "table",
+ "placement": "right",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "asc"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_degraded{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Degraded",
+ "range": true,
+ "refId": "A",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_stale{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Stale",
+ "range": true,
+ "refId": "B",
+ "step": 300
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "sum(ceph_pg_undersized{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "Undersized",
+ "range": true,
+ "refId": "C",
+ "step": 300
+ }
+ ],
+ "title": "Stuck PGs",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 2,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": true,
+ "stacking": {
+ "group": "A",
+ "mode": "none"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "short"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 10,
+ "x": 14,
+ "y": 38
+ },
+ "id": 29,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "list",
+ "placement": "bottom",
+ "showLegend": false
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "sum(irate(ceph_osd_recovery_ops{cluster=~\"$cluster\", }[$interval]))",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "OPS",
+ "refId": "A",
+ "step": 300
+ }
+ ],
+ "title": "Recovery Operations",
+ "type": "timeseries"
+ },
+ {
+ "collapse": false,
+ "collapsed": true,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 44
+ },
+ "id": 30,
+ "panels": [
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": null
+ },
+ "color": {
+ "cardColor": "#b4ff00",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateOranges",
+ "exponent": 0.5,
+ "mode": "opacity"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "scaleDistribution": {
+ "type": "linear"
+ }
+ }
+ }
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 0,
+ "y": 42
+ },
+ "heatmap": { },
+ "hideZeroBuckets": false,
+ "highlightCards": true,
+ "id": 31,
+ "legend": {
+ "show": true
+ },
+ "options": {
+ "calculate": true,
+ "calculation": {
+ "yBuckets": {
+ "mode": "count",
+ "scale": {
+ "log": 2,
+ "type": "log"
+ },
+ "value": "1"
+ }
+ },
+ "cellGap": 2,
+ "cellValues": { },
+ "color": {
+ "exponent": 0.5,
+ "fill": "#b4ff00",
+ "mode": "opacity",
+ "reverse": false,
+ "scale": "exponential",
+ "scheme": "Oranges",
+ "steps": 128
+ },
+ "exemplars": {
+ "color": "rgba(255,0,255,0.7)"
+ },
+ "filterValues": {
+ "le": 1.0000000000000001e-09
+ },
+ "legend": {
+ "show": true
+ },
+ "rowsFrame": {
+ "layout": "auto"
+ },
+ "showValue": "never",
+ "tooltip": {
+ "show": true,
+ "yHistogram": false
+ },
+ "yAxis": {
+ "axisPlacement": "left",
+ "min": "0",
+ "reverse": false,
+ "unit": "ms"
+ }
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_osd_apply_latency_ms{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "OSD Apply Latency Distribution",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "ms",
+ "logBase": 2,
+ "max": null,
+ "min": "0",
+ "show": true,
+ "splitFactor": 1
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": 10
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": null
+ },
+ "color": {
+ "cardColor": "#65c5db",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateOranges",
+ "exponent": 0.5,
+ "mode": "opacity"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "scaleDistribution": {
+ "type": "linear"
+ }
+ }
+ }
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 42
+ },
+ "heatmap": { },
+ "hideZeroBuckets": false,
+ "highlightCards": true,
+ "id": 32,
+ "legend": {
+ "show": true
+ },
+ "options": {
+ "calculate": true,
+ "calculation": {
+ "yBuckets": {
+ "mode": "count",
+ "scale": {
+ "log": 2,
+ "type": "log"
+ }
+ }
+ },
+ "cellGap": 2,
+ "cellValues": { },
+ "color": {
+ "exponent": 0.5,
+ "fill": "#65c5db",
+ "mode": "opacity",
+ "reverse": false,
+ "scale": "exponential",
+ "scheme": "Oranges",
+ "steps": 128
+ },
+ "exemplars": {
+ "color": "rgba(255,0,255,0.7)"
+ },
+ "filterValues": {
+ "le": 1.0000000000000001e-09
+ },
+ "legend": {
+ "show": true
+ },
+ "rowsFrame": {
+ "layout": "auto"
+ },
+ "showValue": "never",
+ "tooltip": {
+ "show": true,
+ "yHistogram": false
+ },
+ "yAxis": {
+ "axisPlacement": "left",
+ "min": "0",
+ "reverse": false,
+ "unit": "ms"
+ }
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "ceph_osd_commit_latency_ms{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "OSD Commit Latency Distribution",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "ms",
+ "logBase": 2,
+ "max": null,
+ "min": "0",
+ "show": true,
+ "splitFactor": 1
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": 10
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": null
+ },
+ "color": {
+ "cardColor": "#806eb7",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateOranges",
+ "exponent": 0.5,
+ "mode": "opacity"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "scaleDistribution": {
+ "type": "linear"
+ }
+ }
+ }
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 0,
+ "y": 50
+ },
+ "heatmap": { },
+ "hideZeroBuckets": false,
+ "highlightCards": true,
+ "id": 33,
+ "legend": {
+ "show": true
+ },
+ "options": {
+ "calculate": true,
+ "calculation": {
+ "yBuckets": {
+ "mode": "count",
+ "scale": {
+ "log": 2,
+ "type": "log"
+ }
+ }
+ },
+ "cellGap": 2,
+ "cellValues": { },
+ "color": {
+ "exponent": 0.5,
+ "fill": "#806eb7",
+ "mode": "opacity",
+ "reverse": false,
+ "scale": "exponential",
+ "scheme": "Oranges",
+ "steps": 128
+ },
+ "exemplars": {
+ "color": "rgba(255,0,255,0.7)"
+ },
+ "filterValues": {
+ "le": 1.0000000000000001e-09
+ },
+ "legend": {
+ "show": true
+ },
+ "rowsFrame": {
+ "layout": "auto"
+ },
+ "showValue": "never",
+ "tooltip": {
+ "show": true,
+ "yHistogram": false
+ },
+ "yAxis": {
+ "axisPlacement": "left",
+ "decimals": 2,
+ "min": "0",
+ "reverse": false,
+ "unit": "ms"
+ }
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[5m]) / rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[5m]) >= 0",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "OSD Read Op Latency Distribution",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "ms",
+ "logBase": 2,
+ "max": null,
+ "min": "0",
+ "show": true,
+ "splitFactor": 1
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": null
+ },
+ "color": {
+ "cardColor": "#f9934e",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateOranges",
+ "exponent": 0.5,
+ "mode": "opacity"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "scaleDistribution": {
+ "type": "linear"
+ }
+ }
+ }
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 50
+ },
+ "heatmap": { },
+ "hideZeroBuckets": false,
+ "highlightCards": true,
+ "id": 34,
+ "legend": {
+ "show": true
+ },
+ "options": {
+ "calculate": true,
+ "calculation": {
+ "yBuckets": {
+ "mode": "count",
+ "scale": {
+ "log": 2,
+ "type": "log"
+ }
+ }
+ },
+ "cellGap": 2,
+ "cellValues": { },
+ "color": {
+ "exponent": 0.5,
+ "fill": "#f9934e",
+ "mode": "opacity",
+ "reverse": false,
+ "scale": "exponential",
+ "scheme": "Oranges",
+ "steps": 128
+ },
+ "exemplars": {
+ "color": "rgba(255,0,255,0.7)"
+ },
+ "filterValues": {
+ "le": 1.0000000000000001e-09
+ },
+ "legend": {
+ "show": true
+ },
+ "rowsFrame": {
+ "layout": "auto"
+ },
+ "showValue": "never",
+ "tooltip": {
+ "show": true,
+ "yHistogram": false
+ },
+ "yAxis": {
+ "axisPlacement": "left",
+ "decimals": 2,
+ "min": "0",
+ "reverse": false,
+ "unit": "ms"
+ }
+ },
+ "pluginVersion": "9.4.7",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[5m]) / rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[5m]) >= 0",
+ "format": "time_series",
+ "instant": false,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "title": "OSD Write Op Latency Distribution",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "ms",
+ "logBase": 2,
+ "max": null,
+ "min": "0",
+ "show": true,
+ "splitFactor": 1
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": false,
+ "stacking": {
+ "group": "A",
+ "mode": "none"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "ms"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 58
+ },
+ "id": 35,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "avg(rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[5m]) / rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[5m]) >= 0)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Read",
+ "refId": "A"
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "avg(rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[5m]) / rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[5m]) >= 0)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Write",
+ "refId": "B"
+ }
+ ],
+ "title": "Recovery Operations",
+ "type": "timeseries"
+ },
+ {
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "axisCenteredZero": false,
+ "axisColorMode": "text",
+ "axisLabel": "",
+ "axisPlacement": "auto",
+ "barAlignment": 0,
+ "drawStyle": "line",
+ "fillOpacity": 10,
+ "gradientMode": "none",
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ },
+ "lineInterpolation": "linear",
+ "lineWidth": 1,
+ "pointSize": 5,
+ "scaleDistribution": {
+ "type": "linear"
+ },
+ "showPoints": "never",
+ "spanNulls": false,
+ "stacking": {
+ "group": "A",
+ "mode": "none"
+ },
+ "thresholdsStyle": {
+ "mode": "off"
+ }
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ },
+ "unit": "ms"
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 58
+ },
+ "id": 36,
+ "interval": "$interval",
+ "options": {
+ "legend": {
+ "calcs": [
+ "lastNotNull",
+ "max"
+ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true
+ },
+ "tooltip": {
+ "mode": "multi",
+ "sort": "none"
+ }
+ },
+ "pluginVersion": "9.1.3",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "expr": "avg(ceph_osd_apply_latency_ms{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "apply",
+ "metric": "ceph_osd_perf_apply_latency_seconds",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "datasource": "$datasource",
+ "expr": "avg(ceph_osd_commit_latency_ms{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "legendFormat": "commit",
+ "metric": "ceph_osd_perf_commit_latency_seconds",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "title": "AVG OSD Apply + Commit Latency",
+ "type": "timeseries"
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "LATENCY",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "collapse": true,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 45
+ },
+ "id": 37,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "columns": [ ],
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "left",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": false,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green"
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Time"
+ },
+ "properties": [
+ {
+ "id": "custom.hidden",
+ "value": true
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 24,
+ "x": 0,
+ "y": 46
+ },
+ "id": 38,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "9.4.7",
+ "styles": "",
+ "targets": [
+ {
+ "datasource": "$datasource",
+ "exemplar": false,
+ "expr": "count by (ceph_version)(ceph_osd_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "OSD Services",
+ "range": false,
+ "refId": "A"
+ },
+ {
+ "datasource": "$datasource",
+ "exemplar": false,
+ "expr": "count by (ceph_version)(ceph_mon_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "Mon Services",
+ "range": false,
+ "refId": "B"
+ },
+ {
+ "datasource": "$datasource",
+ "exemplar": false,
+ "expr": "count by (ceph_version)(ceph_mds_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "MDS Services",
+ "range": false,
+ "refId": "C"
+ },
+ {
+ "datasource": "$datasource",
+ "exemplar": false,
+ "expr": "count by (ceph_version)(ceph_rgw_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "RGW Services",
+ "range": false,
+ "refId": "D"
+ },
+ {
+ "datasource": "$datasource",
+ "exemplar": false,
+ "expr": "count by (ceph_version)(ceph_mgr_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "MGR Services",
+ "range": false,
+ "refId": "E"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Ceph Versions",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": { }
+ },
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": { },
+ "indexByName": { },
+ "renameByName": {
+ "Time": "",
+ "Value #A": "OSD Services",
+ "Value #B": "Mon Services",
+ "Value #C": "MDS Services",
+ "Value #D": "RGW Services",
+ "Value #E": "MGR Services",
+ "ceph_version": "Ceph Version"
+ }
+ }
+ }
+ ],
+ "type": "table"
+ }
+ ],
+ "refresh": "1m",
+ "rows": [ ],
+ "schemaVersion": 38,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "auto": true,
+ "auto_count": 10,
+ "auto_min": "1m",
+ "current": {
+ "text": "$__auto_interval_interval",
+ "value": "$__auto_interval_interval"
+ },
+ "hide": 0,
+ "label": "Interval",
+ "name": "interval",
+ "options": [
+ {
+ "selected": true,
+ "text": "auto",
+ "value": "$__auto_interval_interval"
+ },
+ {
+ "selected": false,
+ "text": "5s",
+ "value": "5s"
+ },
+ {
+ "selected": false,
+ "text": "10s",
+ "value": "10s"
+ },
+ {
+ "selected": false,
+ "text": "30s",
+ "value": "30s"
+ },
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "30m",
+ "value": "30m"
+ },
+ {
+ "selected": false,
+ "text": "1h",
+ "value": "1h"
+ },
+ {
+ "selected": false,
+ "text": "6h",
+ "value": "6h"
+ },
+ {
+ "selected": false,
+ "text": "12h",
+ "value": "12h"
+ },
+ {
+ "selected": false,
+ "text": "1d",
+ "value": "1d"
+ },
+ {
+ "selected": false,
+ "text": "7d",
+ "value": "7d"
+ },
+ {
+ "selected": false,
+ "text": "14d",
+ "value": "14d"
+ },
+ {
+ "selected": false,
+ "text": "30d",
+ "value": "30d"
+ }
+ ],
+ "query": "5s,10s,30s,1m,10m,30m,1h,6h,12h,1d,7d,14d,30d",
+ "refresh": 2,
+ "type": "interval",
+ "valuelabels": { }
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Ceph Cluster - Advanced",
+ "uid": "dn13KBeTv",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/ceph-cluster.json b/roles/kube_prometheus_stack/files/dashboards/ceph-cluster.json
new file mode 100644
index 0000000..3bd150e
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/ceph-cluster.json
@@ -0,0 +1,1432 @@
+{
+ "__requires": [
+ {
+ "type": "grafana",
+ "id": "grafana",
+ "name": "Grafana",
+ "version": "5.0.0"
+ },
+ {
+ "type": "panel",
+ "id": "graph",
+ "name": "Graph",
+ "version": "5.0.0"
+ },
+ {
+ "type": "panel",
+ "id": "heatmap",
+ "name": "Heatmap",
+ "version": "5.0.0"
+ },
+ {
+ "type": "panel",
+ "id": "singlestat",
+ "name": "Singlestat",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "description": "Ceph cluster overview",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "id": null,
+ "iteration": 1525415495309,
+ "links": [],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": true,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 128, 45, 0.9)",
+ "rgba(237, 129, 40, 0.9)",
+ "rgb(255, 0, 0)"
+ ],
+ "datasource": "$datasource",
+ "editable": false,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 0,
+ "y": 0
+ },
+ "hideTimeOverride": true,
+ "id": 21,
+ "interval": "1m",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "ceph_health_status{cluster=~'$cluster'}",
+ "format": "time_series",
+ "instant": true,
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "thresholds": "1,2",
+ "timeFrom": null,
+ "title": "Health Status",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "OK",
+ "value": "0"
+ },
+ {
+ "op": "=",
+ "text": "WARN",
+ "value": "1"
+ },
+ {
+ "op": "=",
+ "text": "ERR",
+ "value": "2"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgb(255, 0, 0)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 0,
+ "datasource": "$datasource",
+ "displayName": "",
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 6,
+ "y": 0
+ },
+ "id": 43,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ }
+ },
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "All",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_osd_metadata{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "All",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "In",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_in{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "In",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Out",
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_in{cluster=~'$cluster'} == bool 0)",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "Out",
+ "refId": "C",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Up",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_up{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Up",
+ "refId": "D",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Down",
+ "crit": 2,
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_osd_up{cluster=~'$cluster'} == bool 0)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Down",
+ "refId": "E",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1
+ }
+ ],
+ "title": "OSDs",
+ "type": "stat",
+ "fieldConfig": {
+ "defaults": {
+ "mappings": [],
+ "thresholds": {
+ "mode": "percentage",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "All"
+ },
+ "properties": [
+ {
+ "id": "color",
+ "value": {
+ "mode": "fixed"
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Out"
+ },
+ "properties": [
+ {
+ "id": "thresholds",
+ "value": {
+ "mode": "percentage",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "orange",
+ "value": 0.1
+ },
+ {
+ "value": 10,
+ "color": "red"
+ }
+ ]
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Down"
+ },
+ "properties": [
+ {
+ "id": "thresholds",
+ "value": {
+ "mode": "percentage",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "orange",
+ "value": 0.1
+ },
+ {
+ "value": 10,
+ "color": "red"
+ }
+ ]
+ }
+ }
+ ]
+ }
+ ]
+ }
+ },
+ {
+ "clusterName": "",
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgba(245, 54, 54, 0.9)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 1,
+ "datasource": "$datasource",
+ "displayName": "",
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 12,
+ "y": 0
+ },
+ "id": 41,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ }
+ },
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "In Quorum",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "sum(ceph_mon_quorum_status{cluster=~'$cluster'})",
+ "format": "time_series",
+ "interval": "",
+ "intervalFactor": 1,
+ "legendFormat": "In Quorum",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Text Only"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Total",
+ "crit": 1,
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mon_quorum_status{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Total",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Text Only",
+ "warn": 2
+ },
+ {
+ "aggregation": "Last",
+ "alias": "MONs out of Quorum",
+ "crit": 1.6,
+ "decimals": 2,
+ "displayAliasType": "Warning / Critical",
+ "displayType": "Annotation",
+ "displayValueWithAlias": "Never",
+ "expr": "count(ceph_mon_quorum_status{cluster=~'$cluster'}) - sum(ceph_mon_quorum_status{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "MONs out of Quorum",
+ "refId": "C",
+ "units": "none",
+ "valueHandler": "Number Threshold",
+ "warn": 1.1
+ }
+ ],
+ "title": "Monitors",
+ "type": "stat"
+ },
+ {
+ "colorMode": "Panel",
+ "colors": {
+ "crit": "rgba(245, 54, 54, 0.9)",
+ "disable": "rgba(128, 128, 128, 0.9)",
+ "ok": "rgba(50, 128, 45, 0.9)",
+ "warn": "rgba(237, 129, 40, 0.9)"
+ },
+ "cornerRadius": 1,
+ "datasource": "$datasource",
+ "displayName": "",
+ "flipCard": false,
+ "flipTime": 5,
+ "fontFormat": "Regular",
+ "gridPos": {
+ "h": 3,
+ "w": 6,
+ "x": 18,
+ "y": 0
+ },
+ "id": 68,
+ "isAutoScrollOnOverflow": false,
+ "isGrayOnNoData": false,
+ "isHideAlertsOnDisable": false,
+ "isIgnoreOKColors": false,
+ "links": [],
+ "options": {
+ "colorMode": "background",
+ "graphMode": "none",
+ "justifyMode": "auto",
+ "orientation": "auto",
+ "reduceOptions": {
+ "calcs": [
+ "lastNotNull"
+ ],
+ "fields": "",
+ "values": false
+ }
+ },
+ "targets": [
+ {
+ "aggregation": "Last",
+ "alias": "Active",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mgr_status{cluster=~'$cluster'} == 1) or vector(0)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "instant": true,
+ "legendFormat": "Active",
+ "refId": "A",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ },
+ {
+ "aggregation": "Last",
+ "alias": "Standby",
+ "decimals": 2,
+ "displayAliasType": "Always",
+ "displayType": "Regular",
+ "displayValueWithAlias": "When Alias Displayed",
+ "expr": "count(ceph_mgr_status{cluster=~'$cluster'} == 0) or vector(0)",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "Standby",
+ "refId": "B",
+ "units": "none",
+ "valueHandler": "Number Threshold"
+ }
+ ],
+ "title": "MGRs",
+ "type": "stat"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "decimals": 2,
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 1,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 6,
+ "x": 0,
+ "y": 6
+ },
+ "id": 47,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_osd_stat_bytes_used{cluster=~'$cluster'})/sum(ceph_osd_stat_bytes{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Used",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "0.7,0.8",
+ "title": "Capacity used",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 9,
+ "x": 6,
+ "y": 6
+ },
+ "id": 53,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "Active",
+ "color": "#508642",
+ "fill": 1,
+ "stack": "A"
+ },
+ {
+ "alias": "Total",
+ "color": "#f9e2d2"
+ },
+ {
+ "alias": "Degraded",
+ "color": "#eab839"
+ },
+ {
+ "alias": "Undersized",
+ "color": "#f9934e"
+ },
+ {
+ "alias": "Inconsistent",
+ "color": "#e24d42"
+ },
+ {
+ "alias": "Down",
+ "color": "#bf1b00"
+ },
+ {
+ "alias": "Inactive",
+ "color": "#bf1b00",
+ "fill": 4,
+ "linewidth": 0,
+ "stack": "A"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(ceph_pg_total{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Total",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(ceph_pg_active{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Active",
+ "refId": "B"
+ },
+ {
+ "expr": "sum(ceph_pg_total{cluster=~'$cluster'} - ceph_pg_active{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Inactive",
+ "refId": "G"
+ },
+ {
+ "expr": "sum(ceph_pg_undersized{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Undersized",
+ "refId": "F"
+ },
+ {
+ "expr": "sum(ceph_pg_degraded{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Degraded",
+ "refId": "C"
+ },
+ {
+ "expr": "sum(ceph_pg_inconsistent{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Inconsistent",
+ "refId": "D"
+ },
+ {
+ "expr": "sum(ceph_pg_down{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Down",
+ "refId": "E"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "PG States",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ms"
+ }
+ },
+ "fill": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 9,
+ "x": 15,
+ "y": 6
+ },
+ "id": 66,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "Avg Apply Latency",
+ "color": "#7eb26d"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "quantile(0.95, ceph_osd_apply_latency_ms{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Apply Latency P_95",
+ "refId": "A"
+ },
+ {
+ "expr": "quantile(0.95, ceph_osd_commit_latency_ms{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Commit Latency P_95",
+ "refId": "B"
+ },
+ {
+ "expr": "avg(ceph_osd_apply_latency_ms{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Avg Apply Latency",
+ "refId": "C"
+ },
+ {
+ "expr": "avg(ceph_osd_commit_latency_ms{cluster=~'$cluster'})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Avg Commit Latency",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD Latencies",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 0,
+ "y": 9
+ },
+ "id": 45,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "Reads",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(irate(ceph_osd_op_w_in_bytes{cluster=~'$cluster'}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(irate(ceph_osd_op_r_out_bytes{cluster=~'$cluster'}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Cluster I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 12,
+ "y": 9
+ },
+ "id": 62,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(deriv(ceph_pool_stored{cluster=~'$cluster'}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "In-/Egress",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": " Egress (-) / Ingress (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": 1
+ },
+ "color": {
+ "cardColor": "rgb(0, 254, 255)",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateBlues",
+ "exponent": 0.5,
+ "min": null,
+ "mode": "spectrum"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 15
+ },
+ "heatmap": {},
+ "highlightCards": true,
+ "id": 55,
+ "legend": {
+ "show": true
+ },
+ "links": [],
+ "span": 12,
+ "targets": [
+ {
+ "expr": "ceph_osd_stat_bytes_used{cluster=~'$cluster'} / ceph_osd_stat_bytes{cluster=~'$cluster'}",
+ "format": "time_series",
+ "interval": "1m",
+ "intervalFactor": 1,
+ "legendFormat": "Util (%)",
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "timeFrom": null,
+ "title": "OSD Capacity Utilization",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": 2,
+ "format": "percentunit",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true,
+ "splitFactor": null
+ },
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": 1
+ },
+ "color": {
+ "cardColor": "#b4ff00",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateBlues",
+ "exponent": 0.5,
+ "mode": "spectrum"
+ },
+ "dataFormat": "timeseries",
+ "datasource": "$datasource",
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 15
+ },
+ "heatmap": {},
+ "highlightCards": true,
+ "id": 59,
+ "legend": {
+ "show": true
+ },
+ "links": [],
+ "targets": [
+ {
+ "expr": "ceph_osd_numpg{cluster=~'$cluster'}",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "#PGs",
+ "refId": "A"
+ }
+ ],
+ "title": "PGs per OSD",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "none",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true,
+ "splitFactor": null
+ },
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ops"
+ }
+ },
+ "fill": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 12,
+ "y": 15
+ },
+ "id": 64,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(irate(ceph_osd_recovery_ops{cluster=~'$cluster'}[1m]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Op/s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Recovery Rate",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": "Recovery Ops/s",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph",
+ "cluster"
+ ],
+ "templating": {
+ "list": [
+ {
+ "hide": 0,
+ "label": null,
+ "name": "datasource",
+ "options": [],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "datasource": {
+ "type": "prometheus",
+ "uid": "${datasource}"
+ },
+ "definition": "label_values(ceph_health_status, cluster)",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [],
+ "query": {
+ "query": "label_values(ceph_health_status, cluster)",
+ "refId": "StandardVariableQuery"
+ },
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "type": "query"
+ },
+ {
+ "auto": true,
+ "auto_count": 10,
+ "auto_min": "1m",
+ "current": {
+ "text": "auto",
+ "value": "$__auto_interval_interval"
+ },
+ "datasource": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "Interval",
+ "multi": false,
+ "name": "interval",
+ "options": [
+ {
+ "selected": true,
+ "text": "auto",
+ "value": "$__auto_interval_interval"
+ },
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "30m",
+ "value": "30m"
+ },
+ {
+ "selected": false,
+ "text": "1h",
+ "value": "1h"
+ },
+ {
+ "selected": false,
+ "text": "6h",
+ "value": "6h"
+ },
+ {
+ "selected": false,
+ "text": "12h",
+ "value": "12h"
+ },
+ {
+ "selected": false,
+ "text": "1d",
+ "value": "1d"
+ },
+ {
+ "selected": false,
+ "text": "7d",
+ "value": "7d"
+ },
+ {
+ "selected": false,
+ "text": "14d",
+ "value": "14d"
+ },
+ {
+ "selected": false,
+ "text": "30d",
+ "value": "30d"
+ }
+ ],
+ "query": "1m,10m,30m,1h,6h,12h,1d,7d,14d,30d",
+ "refresh": 2,
+ "type": "interval"
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Ceph - Cluster",
+ "version": 13
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/host-details.json b/roles/kube_prometheus_stack/files/dashboards/host-details.json
new file mode 100644
index 0000000..ef357d3
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/host-details.json
@@ -0,0 +1,1434 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "singlestat",
+ "name": "Singlestat",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "$ceph_hosts System Overview",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 3,
+ "x": 0,
+ "y": 1
+ },
+ "id": 3,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(sum by (ceph_daemon) (ceph_osd_metadata{cluster=~\"$cluster\", }))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "OSDs",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {
+ "interrupt": "#447EBC",
+ "steal": "#6D1F62",
+ "system": "#890F02",
+ "user": "#3F6833",
+ "wait": "#C15C17"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Shows the CPU breakdown. When multiple servers are selected, only the first host's cpu data is shown",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "percent"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 10,
+ "w": 6,
+ "x": 3,
+ "y": 1
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (mode) (\n rate(node_cpu{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[$__rate_interval]) or\n rate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", mode=~\"(irq|nice|softirq|steal|system|user|iowait)\"}[$__rate_interval])\n) / (\n scalar(\n sum(rate(node_cpu{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_cpu_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]))\n ) * 100\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{mode}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Utilization",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": "% Utilization",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+ "Available": "#508642",
+ "Free": "#508642",
+ "Total": "#bf1b00",
+ "Used": "#bf1b00",
+ "total": "#bf1b00",
+ "used": "#0a50a1"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "bytes"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 10,
+ "w": 6,
+ "x": 9,
+ "y": 1
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "total",
+ "color": "#bf1b00",
+ "fill": 0,
+ "linewidth": 2,
+ "stack": false
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_memory_MemFree{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Free",
+ "refId": "A"
+ },
+ {
+ "expr": "node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "total",
+ "refId": "B"
+ },
+ {
+ "expr": "(\n node_memory_Cached{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) + (\n node_memory_Buffers{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) + (\n node_memory_Slab{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "buffers/cache",
+ "refId": "C"
+ },
+ {
+ "expr": "(\n node_memory_MemTotal{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemTotal_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n) - (\n (\n node_memory_MemFree{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_MemFree_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) + (\n node_memory_Cached{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Cached_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) + (\n node_memory_Buffers{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Buffers_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n ) +\n (\n node_memory_Slab{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"} or\n node_memory_Slab_bytes{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}\n )\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "used",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "RAM Usage",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": "RAM used",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Show the network load (rx,tx) across all interfaces (excluding loopback 'lo')",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "decbytes"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 10,
+ "w": 6,
+ "x": 15,
+ "y": 1
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (device) (\n rate(\n node_network_receive_bytes{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_receive_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[$__rate_interval]\n )\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.rx",
+ "refId": "A"
+ },
+ {
+ "expr": "sum by (device) (\n rate(node_network_transmit_bytes{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_transmit_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\",device!=\"lo\"}[$__rate_interval])\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.tx",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Load",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "decbytes",
+ "label": "Send (-) / Receive (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "pps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 5,
+ "w": 3,
+ "x": 21,
+ "y": 1
+ },
+ "id": 7,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_drop{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_network_receive_drop_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.rx",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(node_network_transmit_drop{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_network_transmit_drop_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.tx",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network drop rate",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "pps",
+ "label": "Send (-) / Receive (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Each OSD consists of a Journal/WAL partition and a data partition. The RAW Capacity shown is the sum of the data partitions across all OSDs on the selected OSD hosts.",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 3,
+ "x": 0,
+ "y": 6
+ },
+ "id": 8,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(\n ceph_osd_stat_bytes{cluster=~\"$cluster\", } and\n on (ceph_daemon) ceph_disk_occupation{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", cluster=~\"$cluster\", }\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Raw Capacity",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "pps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 5,
+ "w": 3,
+ "x": 21,
+ "y": 6
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_errs{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_network_receive_errs_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.rx",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(node_network_transmit_errs{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_network_transmit_errs_total{instance=~\"$ceph_hosts([\\\\\\\\.:].*)?\"}[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.tx",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network error rate",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "pps",
+ "label": "Send (-) / Receive (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 11
+ },
+ "id": 10,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "OSD Disk Performance Statistics",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "For any OSD devices on the host, this chart shows the iops per physical device. Each device is shown by it's name and corresponding OSD id value",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ops"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 0,
+ "y": 12
+ },
+ "id": 11,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*reads/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n (\n rate(node_disk_writes_completed{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_disk_writes_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{cluster=~\"$cluster\", }, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) writes",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(\n (\n rate(node_disk_reads_completed{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_disk_reads_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{cluster=~\"$cluster\", },\"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "For OSD hosts, this chart shows the disk bandwidth (read bytes/sec + write bytes/sec) of the physical OSD device. Each device is shown by device name, and corresponding OSD id",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 12,
+ "y": 12
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*read/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n (\n rate(node_disk_bytes_written{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_disk_written_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device)\n group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human{cluster=~\"$cluster\", }, \"device\", \"$1\", \"device\", \"/dev/(.*)\"),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) write",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(\n (\n rate(node_disk_bytes_read{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) or\n rate(node_disk_read_bytes_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])\n ),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") * on(instance, device)\n group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human{cluster=~\"$cluster\", }, \"device\", \"$1\", \"device\", \"/dev/(.*)\"),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Throughput by Disk",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "For OSD hosts, this chart shows the latency at the physical drive. Each drive is shown by device name, with it's corresponding OSD id",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "s"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 0,
+ "y": 21
+ },
+ "id": 13,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max by(instance, device) (label_replace(\n (rate(node_disk_write_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])) /\n clamp_min(rate(node_disk_writes_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]), 0.001) or\n (rate(node_disk_read_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval])) /\n clamp_min(rate(node_disk_reads_completed_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]), 0.001),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}})",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Show disk utilization % (util) of any OSD devices on the host by the physical device name and associated OSD id.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "percent"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 12,
+ "y": 21
+ },
+ "id": 14,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n (\n (rate(node_disk_io_time_ms{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) / 10) or\n rate(node_disk_io_time_seconds_total{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\"}[$__rate_interval]) * 100\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(ceph_disk_occupation_human{instance=~\"($ceph_hosts)([\\\\\\\\.:].*)?\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}})",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk utilization",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": "%Util",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "instance"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Instance"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Slow Ops"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 6,
+ "x": 0,
+ "y": 30
+ },
+ "id": 15,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n (sum by (instance)(ceph_daemon_health_metrics{type=\"SLOW_OPS\", ceph_daemon=~\"osd.*\", cluster=~\"$cluster\", }))\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Top Slow Ops per Host",
+ "transformations": [
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": {
+ "Time": true,
+ "cluster": true
+ },
+ "includeByName": { },
+ "indexByName": { },
+ "renameByName": { }
+ }
+ }
+ ],
+ "type": "table"
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin",
+ "overview"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "ceph_hosts",
+ "options": [ ],
+ "query": "label_values({__name__=~\"ceph_.+_metadata\", cluster=~\"$cluster\", }, hostname)",
+ "refresh": 1,
+ "regex": "([^.]*).*",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Host Details",
+ "uid": "rtOg0AiWz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/hosts-overview.json b/roles/kube_prometheus_stack/files/dashboards/hosts-overview.json
new file mode 100644
index 0000000..f1cd4c4
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/hosts-overview.json
@@ -0,0 +1,894 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "singlestat",
+ "name": "Singlestat",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(sum by (hostname) (ceph_osd_metadata{job=~\"$job\"}))",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "OSD Hosts",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Average CPU busy across all hosts (OSD, RGW, MON etc) within the cluster",
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 4,
+ "y": 0
+ },
+ "id": 3,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "avg(1 - (\n avg by(instance) (\n rate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[$__rate_interval]) or\n rate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[$__rate_interval])\n )\n))\n",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "AVG CPU Busy",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Average Memory Usage across all hosts in the cluster (excludes buffer/cache usage)",
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 8,
+ "y": 0
+ },
+ "id": 4,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "avg ((\n (\n node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) - ((\n node_memory_MemFree{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemFree_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}) +\n (\n node_memory_Cached{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Cached_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) + (\n node_memory_Buffers{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Buffers_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n ) + (\n node_memory_Slab{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_Slab_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}\n )\n )\n) / (\n node_memory_MemTotal{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"} or\n node_memory_MemTotal_bytes{instance=~\"($osd_hosts|$rgw_hosts|$mon_hosts|$mds_hosts).*\"}\n))\n",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "AVG RAM Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "IOPS Load at the device as reported by the OS on all OSD hosts",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 12,
+ "y": 0
+ },
+ "id": 5,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum ((\n rate(node_disk_reads_completed{instance=~\"($osd_hosts).*\"}[$__rate_interval]) or\n rate(node_disk_reads_completed_total{instance=~\"($osd_hosts).*\"}[$__rate_interval])\n) + (\n rate(node_disk_writes_completed{instance=~\"($osd_hosts).*\"}[$__rate_interval]) or\n rate(node_disk_writes_completed_total{instance=~\"($osd_hosts).*\"}[$__rate_interval])\n))\n",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Physical IOPS",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Average Disk utilization for all OSD data devices (i.e. excludes journal/WAL)",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 16,
+ "y": 0
+ },
+ "id": 6,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "avg (\n label_replace(\n (rate(node_disk_io_time_ms[$__rate_interval]) / 10 ) or\n (rate(node_disk_io_time_seconds_total[$__rate_interval]) * 100),\n \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n ) * on(instance, device) group_left(ceph_daemon) label_replace(\n label_replace(\n ceph_disk_occupation_human{job=~\"$job\", instance=~\"($osd_hosts).*\"},\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^.:]*).*\"\n )\n)\n",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "AVG Disk Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Total send/receive network load across all hosts in the ceph cluster",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 5,
+ "w": 4,
+ "x": 20,
+ "y": 0
+ },
+ "id": 7,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum (\n (\n rate(node_network_receive_bytes{instance=~\"($osd_hosts|mon_hosts|mds_hosts|rgw_hosts).*\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_receive_bytes_total{instance=~\"($osd_hosts|mon_hosts|mds_hosts|rgw_hosts).*\",device!=\"lo\"}[$__rate_interval])\n ) unless on (device, instance)\n label_replace((bonding_slaves > 0), \"device\", \"$1\", \"master\", \"(.+)\")\n) +\nsum (\n (\n rate(node_network_transmit_bytes{instance=~\"($osd_hosts|mon_hosts|mds_hosts|rgw_hosts).*\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_transmit_bytes_total{instance=~\"($osd_hosts|mon_hosts|mds_hosts|rgw_hosts).*\",device!=\"lo\"}[$__rate_interval])\n ) unless on (device, instance)\n label_replace((bonding_slaves > 0), \"device\", \"$1\", \"master\", \"(.+)\")\n)\n",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Network Load",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Show the top 10 busiest hosts by cpu",
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 0,
+ "y": 5
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "topk(10,\n 100 * (\n 1 - (\n avg by(instance) (\n rate(node_cpu_seconds_total{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[$__rate_interval]) or\n rate(node_cpu{mode='idle',instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\"}[$__rate_interval])\n )\n )\n )\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "CPU Busy - Top 10 Hosts",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Top 10 hosts by network load",
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 12,
+ "y": 5
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "topk(10, (sum by(instance) (\n(\n rate(node_network_receive_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_receive_bytes_total{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\",device!=\"lo\"}[$__rate_interval])\n) +\n(\n rate(node_network_transmit_bytes{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\",device!=\"lo\"}[$__rate_interval]) or\n rate(node_network_transmit_bytes_total{instance=~\"($osd_hosts|$mon_hosts|$mds_hosts|$rgw_hosts).*\",device!=\"lo\"}[$__rate_interval])\n) unless on (device, instance)\n label_replace((bonding_slaves > 0), \"device\", \"$1\", \"master\", \"(.+)\"))\n))\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Network Load - Top 10 Hosts",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": ".+",
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": true,
+ "label": "cluster",
+ "multi": true,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_osd_metadata, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": ".+",
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": "job",
+ "multi": true,
+ "name": "job",
+ "options": [ ],
+ "query": "label_values(ceph_osd_metadata{}, job)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "osd_hosts",
+ "options": [ ],
+ "query": "label_values(ceph_disk_occupation{job=~\"$job\"}, exported_instance)",
+ "refresh": 1,
+ "regex": "([^.]*).*",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "mon_hosts",
+ "options": [ ],
+ "query": "label_values(ceph_mon_metadata{job=~\"$job\"}, ceph_daemon)",
+ "refresh": 1,
+ "regex": "mon.(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "mds_hosts",
+ "options": [ ],
+ "query": "label_values(ceph_mds_inodes{job=~\"$job\"}, ceph_daemon)",
+ "refresh": 1,
+ "regex": "mds.(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "rgw_hosts",
+ "options": [ ],
+ "query": "label_values(ceph_rgw_metadata{job=~\"$job\"}, ceph_daemon)",
+ "refresh": 1,
+ "regex": "rgw.(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Host Overview",
+ "uid": "y0KGL0iZz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/osd-device-details.json b/roles/kube_prometheus_stack/files/dashboards/osd-device-details.json
new file mode 100644
index 0000000..60f1ecc
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/osd-device-details.json
@@ -0,0 +1,914 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "OSD Performance",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "s"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 1
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_osd_op_r_latency_sum{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "read",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_osd_op_w_latency_sum{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "write",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$osd Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 1
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "Reads",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_osd_op_r{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_osd_op_w{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$osd R/W IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "bytes"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 12,
+ "y": 1
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "Read Bytes",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_osd_op_r_out_bytes{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Read Bytes",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_osd_op_w_in_bytes{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Write Bytes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$osd R/W Bytes",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 10
+ },
+ "id": 6,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "Physical Device Performance",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "s"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 11
+ },
+ "id": 7,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*Reads/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "(\n label_replace(\n rate(node_disk_read_time_seconds_total[$__rate_interval]) /\n rate(node_disk_reads_completed_total[$__rate_interval]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n ) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}/{{device}} Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "(\n label_replace(\n rate(node_disk_write_time_seconds_total[$__rate_interval]) /\n rate(node_disk_writes_completed_total[$__rate_interval]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\") and on (instance, device)\n label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n )\n )\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}/{{device}} Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device Latency for $osd",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 11
+ },
+ "id": 8,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*Reads/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n rate(node_disk_writes_completed_total[$__rate_interval]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}} on {{instance}} Writes",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(\n rate(node_disk_reads_completed_total[$__rate_interval]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}} on {{instance}} Reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device R/W IOPS for $osd",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 12,
+ "y": 11
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "/.*Reads/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n rate(node_disk_read_bytes_total[$__rate_interval]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(\n rate(node_disk_written_bytes_total[$__rate_interval]), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", },\n \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device R/W Bytes for $osd",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "percentunit"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 18,
+ "y": 11
+ },
+ "id": 10,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(\n rate(node_disk_io_time_seconds_total[$__rate_interval]),\n \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n) and on (instance, device) label_replace(\n label_replace(\n ceph_disk_occupation_human{ceph_daemon=~\"$osd\", cluster=~\"$cluster\", }, \"device\", \"$1\", \"device\", \"/dev/(.*)\"\n ), \"instance\", \"$1\", \"instance\", \"([^:.]*).*\"\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}} on {{instance}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device Util% for $osd",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "OSD",
+ "multi": false,
+ "name": "osd",
+ "options": [ ],
+ "query": "label_values(ceph_osd_metadata{cluster=~\"$cluster\", }, ceph_daemon)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-3h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "OSD device details",
+ "uid": "CrAHE0iZz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/osds-overview.json b/roles/kube_prometheus_stack/files/dashboards/osds-overview.json
new file mode 100644
index 0000000..948f0d7
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/osds-overview.json
@@ -0,0 +1,1339 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.0.0"
+ },
+ {
+ "id": "grafana-piechart-panel",
+ "name": "Pie Chart",
+ "type": "panel",
+ "version": "1.3.3"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "table",
+ "name": "Table",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "aliasColors": {
+ "@95%ile": "#e0752d"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ms"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg (\n rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) * 1000\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "AVG read",
+ "refId": "A"
+ },
+ {
+ "expr": "max(\n rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) * 1000\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "MAX read",
+ "refId": "B"
+ },
+ {
+ "expr": "quantile(0.95,\n (\n rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[$__rate_interval])\n * 1000\n )\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "@95%ile",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD Read Latencies",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ }
+ ]
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "ceph_daemon"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "OSD ID"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Latency (ms)"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 8,
+ "y": 0
+ },
+ "id": 3,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n (sort(\n (\n rate(ceph_osd_op_r_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_r_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) *\n 1000\n )\n ))\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Highest READ Latencies",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ },
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": {
+ "Time": true,
+ "cluster": true
+ },
+ "includeByName": { },
+ "indexByName": { },
+ "renameByName": { }
+ }
+ }
+ ],
+ "type": "table"
+ },
+ {
+ "aliasColors": {
+ "@95%ile write": "#e0752d"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ms"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 12,
+ "y": 0
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(\n rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[$__rate_interval])\n * 1000\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "AVG write",
+ "refId": "A"
+ },
+ {
+ "expr": "max(\n rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) *\n 1000\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "MAX write",
+ "refId": "B"
+ },
+ {
+ "expr": "quantile(0.95, (\n rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) *\n 1000\n))\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "@95%ile write",
+ "refId": "C"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD Write Latencies",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ }
+ ]
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "description": "This table shows the osd's that are delivering the 10 highest write latencies within the cluster",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "ceph_daemon"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "OSD ID"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Latency (ms)"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "mappings",
+ "value": [
+ {
+ "options": {
+ "NaN": {
+ "index": 0,
+ "text": "0.00"
+ }
+ },
+ "type": "value"
+ }
+ ]
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 20,
+ "y": 0
+ },
+ "id": 5,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n (sort(\n (rate(ceph_osd_op_w_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n on (ceph_daemon) rate(ceph_osd_op_w_latency_count{cluster=~\"$cluster\", }[$__rate_interval]) *\n 1000)\n ))\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Highest WRITE Latencies",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ },
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": {
+ "Time": true,
+ "cluster": true
+ },
+ "includeByName": { },
+ "indexByName": { },
+ "renameByName": { }
+ }
+ }
+ ],
+ "type": "table"
+ },
+ {
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ }
+ },
+ "mappings": [ ]
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 0,
+ "y": 8
+ },
+ "id": 6,
+ "options": {
+ "displayLabels": [
+ "percent"
+ ],
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true,
+ "values": [
+ "percent",
+ "value"
+ ]
+ },
+ "pieType": "pie",
+ "reduceOptions": { },
+ "tooltip": {
+ "mode": "single",
+ "sort": "none"
+ }
+ },
+ "targets": [
+ {
+ "expr": "count by (device_class) (ceph_osd_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device_class}}",
+ "refId": "A"
+ }
+ ],
+ "title": "OSD Types Summary",
+ "type": "piechart"
+ },
+ {
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ }
+ },
+ "mappings": [ ]
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 4,
+ "y": 8
+ },
+ "id": 7,
+ "options": {
+ "displayLabels": [
+ "percent"
+ ],
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true,
+ "values": [
+ "percent",
+ "value"
+ ]
+ },
+ "pieType": "pie",
+ "reduceOptions": { },
+ "tooltip": {
+ "mode": "single",
+ "sort": "none"
+ }
+ },
+ "targets": [
+ {
+ "expr": "count(ceph_bluefs_wal_total_bytes{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "bluestore",
+ "refId": "A"
+ },
+ {
+ "expr": "absent(ceph_bluefs_wal_total_bytes{cluster=~\"$cluster\", }) * count(ceph_osd_metadata{cluster=~\"$cluster\", })",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "filestore",
+ "refId": "B"
+ }
+ ],
+ "title": "OSD Objectstore Types",
+ "type": "piechart"
+ },
+ {
+ "datasource": "$datasource",
+ "description": "The pie chart shows the various OSD sizes used within the cluster",
+ "fieldConfig": {
+ "defaults": {
+ "color": {
+ "mode": "palette-classic"
+ },
+ "custom": {
+ "hideFrom": {
+ "legend": false,
+ "tooltip": false,
+ "viz": false
+ }
+ },
+ "mappings": [ ]
+ },
+ "overrides": [ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 8,
+ "y": 8
+ },
+ "id": 8,
+ "options": {
+ "displayLabels": [
+ "percent"
+ ],
+ "legend": {
+ "calcs": [ ],
+ "displayMode": "table",
+ "placement": "bottom",
+ "showLegend": true,
+ "values": [
+ "percent",
+ "value"
+ ]
+ },
+ "pieType": "pie",
+ "reduceOptions": { },
+ "tooltip": {
+ "mode": "single",
+ "sort": "none"
+ }
+ },
+ "targets": [
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } < 1099511627776)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<1TB",
+ "refId": "A"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 1099511627776 < 2199023255552)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<2TB",
+ "refId": "B"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 2199023255552 < 3298534883328)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<3TB",
+ "refId": "C"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 3298534883328 < 4398046511104)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<4TB",
+ "refId": "D"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 4398046511104 < 6597069766656)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<6TB",
+ "refId": "E"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 6597069766656 < 8796093022208)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<8TB",
+ "refId": "F"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 8796093022208 < 10995116277760)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<10TB",
+ "refId": "G"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 10995116277760 < 13194139533312)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<12TB",
+ "refId": "H"
+ },
+ {
+ "expr": "count(ceph_osd_stat_bytes{cluster=~\"$cluster\", } >= 13194139533312)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "<12TB+",
+ "refId": "I"
+ }
+ ],
+ "title": "OSD Size Summary",
+ "type": "piechart"
+ },
+ {
+ "aliasColors": { },
+ "bars": true,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 8,
+ "x": 12,
+ "y": 8
+ },
+ "id": 9,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "ceph_osd_numpg{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "PGs per OSD",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Distribution of PGs per OSD",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": 20,
+ "mode": "histogram",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "# of OSDs",
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ }
+ ]
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "This gauge panel shows onode Hits ratio to help determine if increasing RAM per OSD could help improve the performance of the cluster",
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 1,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 4,
+ "x": 20,
+ "y": 8
+ },
+ "id": 10,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_bluestore_onode_hits{cluster=~\"$cluster\", }) / (\n sum(ceph_bluestore_onode_hits{cluster=~\"$cluster\", }) +\n sum(ceph_bluestore_onode_misses{cluster=~\"$cluster\", })\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": ".75",
+ "title": "OSD onode Hits Ratio",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "collapse": false,
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 16
+ },
+ "id": 11,
+ "panels": [ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": true,
+ "title": "R/W Profile",
+ "titleSize": "h6",
+ "type": "row"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Show the read/write workload profile overtime",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 24,
+ "x": 0,
+ "y": 17
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "round(sum(rate(ceph_pool_rd{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "round(sum(rate(ceph_pool_wr{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Read/Write Profile",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "description": "This table shows the 10 OSDs with the highest number of slow ops",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "ceph_daemon"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "OSD ID"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Slow Ops"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 8,
+ "w": 5,
+ "x": 0,
+ "y": 25
+ },
+ "id": 13,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n (ceph_daemon_health_metrics{type=\"SLOW_OPS\", ceph_daemon=~\"osd.*\"})\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Top Slow Ops",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ },
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": {
+ "Time": true,
+ "__name__": true,
+ "cluster": true,
+ "instance": true,
+ "job": true,
+ "type": true
+ },
+ "includeByName": { },
+ "indexByName": { },
+ "renameByName": { }
+ }
+ }
+ ],
+ "type": "table"
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "OSD Overview",
+ "uid": "lo02I1Aiz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/pool-detail.json b/roles/kube_prometheus_stack/files/dashboards/pool-detail.json
new file mode 100644
index 0000000..5e5bf6e
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/pool-detail.json
@@ -0,0 +1,724 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "singlestat",
+ "name": "Singlestat",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 1,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 7,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(ceph_pool_stored{cluster=~\"$cluster\", } / (ceph_pool_stored{cluster=~\"$cluster\", } + ceph_pool_max_avail{cluster=~\"$cluster\", })) *\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": ".7,.8",
+ "title": "Capacity used",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": 100,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Time till pool is full assuming the average fill rate of the last 6 hours",
+ "format": "s",
+ "gauge": {
+ "maxValue": false,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 5,
+ "x": 7,
+ "y": 0
+ },
+ "id": 3,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": ""
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(ceph_pool_max_avail{cluster=~\"$cluster\", } / deriv(ceph_pool_stored{cluster=~\"$cluster\", }[6h])) *\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", } > 0\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "current",
+ "title": "Time till full",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": false
+ },
+ {
+ "aliasColors": {
+ "read_op_per_sec": "#3F6833",
+ "write_op_per_sec": "#E5AC0E"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ops"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 0
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "deriv(ceph_pool_objects{cluster=~\"$cluster\", }[1m]) *\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Objects per second",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$pool_name Object Ingress/Egress",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": "Objects out(-) / in(+) ",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+ "read_op_per_sec": "#3F6833",
+ "write_op_per_sec": "#E5AC0E"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "iops"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 7
+ },
+ "id": 5,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "reads",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_pool_rd{cluster=~\"$cluster\", }[$__rate_interval]) *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "reads",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_pool_wr{cluster=~\"$cluster\", }[$__rate_interval]) *\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$pool_name Client IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "iops",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+ "read_op_per_sec": "#3F6833",
+ "write_op_per_sec": "#E5AC0E"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 7
+ },
+ "id": 6,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [
+ {
+ "alias": "reads",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_pool_rd_bytes{cluster=~\"$cluster\", }[$__rate_interval]) +\n on(pool_id) group_left(instance, name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "reads",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_pool_wr_bytes{cluster=~\"$cluster\", }[$__rate_interval]) +\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$pool_name Client Throughput",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {
+ "read_op_per_sec": "#3F6833",
+ "write_op_per_sec": "#E5AC0E"
+ },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 14
+ },
+ "id": 7,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "ceph_pool_objects{cluster=~\"$cluster\", } *\n on(pool_id) group_left(instance,name) ceph_pool_metadata{name=~\"$pool_name\", cluster=~\"$cluster\", }\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Number of Objects",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$pool_name Objects",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "Objects",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 22,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Pool Name",
+ "multi": false,
+ "name": "pool_name",
+ "options": [ ],
+ "query": "label_values(ceph_pool_metadata{cluster=~\"$cluster\", }, name)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Ceph Pool Details",
+ "uid": "-xyV8KCiz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/pool-overview.json b/roles/kube_prometheus_stack/files/dashboards/pool-overview.json
new file mode 100644
index 0000000..fa32b33
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/pool-overview.json
@@ -0,0 +1,1691 @@
+{
+ "__inputs": [ ],
+ "__requires": [ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(ceph_pool_metadata{cluster=~\"$cluster\", })",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Pools",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Count of the pools that have compression enabled",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 3,
+ "y": 0
+ },
+ "id": 3,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "count(ceph_pool_metadata{compression_mode!=\"none\", cluster=~\"$cluster\", })",
+ "format": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Pools with Compression",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Total raw capacity available to the cluster",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 6,
+ "y": 0
+ },
+ "id": 4,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_osd_stat_bytes{cluster=~\"$cluster\", })",
+ "format": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Total Raw Capacity",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Total raw capacity consumed by user data and associated overheads (metadata + redundancy)",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 9,
+ "y": 0
+ },
+ "id": 5,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_pool_bytes_used{cluster=~\"$cluster\", })",
+ "format": "",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Raw Capacity Consumed",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Total of client data stored in the cluster",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 12,
+ "y": 0
+ },
+ "id": 6,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_pool_stored{cluster=~\"$cluster\", })",
+ "format": "",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Logical Stored ",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "A compression saving is determined as the data eligible to be compressed minus the capacity used to store the data after compression",
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 15,
+ "y": 0
+ },
+ "id": 7,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(\n ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } -\n ceph_pool_compress_bytes_used{cluster=~\"$cluster\", }\n)\n",
+ "format": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Compression Savings",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "Indicates how suitable the data is within the pools that are/have been enabled for compression - averaged across all pools holding compressed data",
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 18,
+ "y": 0
+ },
+ "id": 8,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "(\n sum(ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } > 0) /\n sum(ceph_pool_stored_raw{cluster=~\"$cluster\", } and ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } > 0)\n) * 100\n",
+ "format": "table",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Compression Eligibility",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": "$datasource",
+ "description": "This factor describes the average ratio of data eligible to be compressed divided by the data actually stored. It does not account for data written that was ineligible for compression (too small, or compression yield too low)",
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 3,
+ "x": 21,
+ "y": 0
+ },
+ "id": 9,
+ "interval": null,
+ "links": [ ],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(\n ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } > 0)\n / sum(ceph_pool_compress_bytes_used{cluster=~\"$cluster\", } > 0\n)\n",
+ "format": "",
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "",
+ "title": "Compression Factor",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "auto",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Time"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "instance"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "job"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "name"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Pool Name"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "pool_id"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Pool ID"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #A"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Compression Factor"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #D"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "% Used"
+ },
+ {
+ "id": "unit",
+ "value": "percentunit"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.cellOptions",
+ "value": {
+ "type": "color-text"
+ }
+ },
+ {
+ "id": "thresholds",
+ "value": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 70
+ },
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": 85
+ }
+ ]
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #B"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Usable Free"
+ },
+ {
+ "id": "unit",
+ "value": "bytes"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #C"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Compression Eligibility"
+ },
+ {
+ "id": "unit",
+ "value": "percent"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #E"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Compression Savings"
+ },
+ {
+ "id": "unit",
+ "value": "bytes"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #F"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Growth (5d)"
+ },
+ {
+ "id": "unit",
+ "value": "bytes"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.cellOptions",
+ "value": {
+ "type": "color-text"
+ }
+ },
+ {
+ "id": "thresholds",
+ "value": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "rgba(245, 54, 54, 0.9)",
+ "value": null
+ },
+ {
+ "color": "rgba(237, 129, 40, 0.89)",
+ "value": 70
+ },
+ {
+ "color": "rgba(50, 172, 45, 0.97)",
+ "value": 85
+ }
+ ]
+ }
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #G"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "IOPS"
+ },
+ {
+ "id": "unit",
+ "value": "none"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #H"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Bandwidth"
+ },
+ {
+ "id": "unit",
+ "value": "Bps"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "__name__"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "type"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "compression_mode"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "description"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Type"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #J"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Stored"
+ },
+ {
+ "id": "unit",
+ "value": "bytes"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #I"
+ },
+ "properties": [
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value #K"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Compression"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 24,
+ "x": 0,
+ "y": 3
+ },
+ "id": 10,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "(\n ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } /\n ceph_pool_compress_bytes_used{cluster=~\"$cluster\", } > 0\n) and on(pool_id) (\n (\n (ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } > 0) /\n ceph_pool_stored_raw{cluster=~\"$cluster\", }\n ) * 100 > 0.5\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "A",
+ "refId": "A"
+ },
+ {
+ "expr": "ceph_pool_max_avail{cluster=~\"$cluster\", } *\n on(pool_id) group_left(name) ceph_pool_metadata{cluster=~\"$cluster\", }\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "B",
+ "refId": "B"
+ },
+ {
+ "expr": "(\n (ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } > 0) /\n ceph_pool_stored_raw{cluster=~\"$cluster\", }\n) * 100\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "C",
+ "refId": "C"
+ },
+ {
+ "expr": "ceph_pool_percent_used{cluster=~\"$cluster\", } *\n on(pool_id) group_left(name) ceph_pool_metadata{cluster=~\"$cluster\", }\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "D",
+ "refId": "D"
+ },
+ {
+ "expr": "ceph_pool_compress_under_bytes{cluster=~\"$cluster\", } -\n ceph_pool_compress_bytes_used{cluster=~\"$cluster\", } > 0\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "E",
+ "refId": "E"
+ },
+ {
+ "expr": "delta(ceph_pool_stored{cluster=~\"$cluster\", }[5d])",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "F",
+ "refId": "F"
+ },
+ {
+ "expr": "rate(ceph_pool_rd{cluster=~\"$cluster\", }[$__rate_interval])\n + rate(ceph_pool_wr{cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "G",
+ "refId": "G"
+ },
+ {
+ "expr": "rate(ceph_pool_rd_bytes{cluster=~\"$cluster\", }[$__rate_interval]) +\n rate(ceph_pool_wr_bytes{cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "H",
+ "refId": "H"
+ },
+ {
+ "expr": "ceph_pool_metadata{cluster=~\"$cluster\", }",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "I",
+ "refId": "I"
+ },
+ {
+ "expr": "ceph_pool_stored{cluster=~\"$cluster\", } * on(pool_id) group_left ceph_pool_metadata{cluster=~\"$cluster\", }",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "J",
+ "refId": "J"
+ },
+ {
+ "expr": "ceph_pool_metadata{compression_mode!=\"none\", cluster=~\"$cluster\", }",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "K",
+ "refId": "K"
+ },
+ {
+ "expr": "",
+ "format": "",
+ "intervalFactor": "",
+ "legendFormat": "L",
+ "refId": "L"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Pool Overview",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": { }
+ },
+ {
+ "id": "seriesToRows",
+ "options": { }
+ },
+ {
+ "id": "organize",
+ "options": {
+ "excludeByName": {
+ "Time": true,
+ "Value #A": true,
+ "Value #B": false,
+ "Value #C": true,
+ "Value #D": false,
+ "Value #E": true,
+ "Value #I": true,
+ "Value #K": true,
+ "__name__": true,
+ "cluster": true,
+ "compression_mode": true,
+ "instance": true,
+ "job": true,
+ "pool_id": true,
+ "type": true
+ },
+ "includeByName": { },
+ "indexByName": { },
+ "renameByName": { }
+ }
+ }
+ ],
+ "type": "table"
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "This chart shows the sum of read and write IOPS from all clients by pool",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 0,
+ "y": 9
+ },
+ "id": 11,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "topk($topk,\n round(\n (\n rate(ceph_pool_rd{cluster=~\"$cluster\", }[$__rate_interval]) +\n rate(ceph_pool_wr{cluster=~\"$cluster\", }[$__rate_interval])\n ), 1\n ) * on(pool_id) group_left(instance,name) ceph_pool_metadata{cluster=~\"$cluster\", })\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}} ",
+ "refId": "A"
+ },
+ {
+ "expr": "topk($topk,\n rate(ceph_pool_wr{cluster=~\"$cluster\", }[$__rate_interval]) +\n on(pool_id) group_left(instance,name) ceph_pool_metadata{cluster=~\"$cluster\", }\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}} - write",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Top $topk Client IOPS by Pool",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "IOPS",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "The chart shows the sum of read and write bytes from all clients, by pool",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 8,
+ "w": 12,
+ "x": 12,
+ "y": 9
+ },
+ "id": 12,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "topk($topk,\n (\n rate(ceph_pool_rd_bytes{cluster=~\"$cluster\", }[$__rate_interval]) +\n rate(ceph_pool_wr_bytes{cluster=~\"$cluster\", }[$__rate_interval])\n ) * on(pool_id) group_left(instance, name) ceph_pool_metadata{cluster=~\"$cluster\", }\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Top $topk Client Bandwidth by Pool",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Throughput",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "Historical view of capacity usage, to help identify growth and trends in pool consumption",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "bytes"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 24,
+ "x": 0,
+ "y": 17
+ },
+ "id": 13,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "ceph_pool_bytes_used{cluster=~\"$cluster\", } * on(pool_id) group_right ceph_pool_metadata{cluster=~\"$cluster\", }",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Pool Capacity Usage (RAW)",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": "Capacity Used",
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 22,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "15",
+ "value": "15"
+ },
+ "hide": 0,
+ "includeAll": false,
+ "label": "TopK",
+ "multi": false,
+ "name": "topk",
+ "options": [
+ {
+ "text": "15",
+ "value": "15"
+ }
+ ],
+ "query": "15",
+ "refresh": 0,
+ "type": "custom"
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "Ceph Pools Overview",
+ "uid": "z99hzWtmk",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/rbd-details.json b/roles/kube_prometheus_stack/files/dashboards/rbd-details.json
new file mode 100644
index 0000000..500c51f
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/rbd-details.json
@@ -0,0 +1,465 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.3.3"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Detailed Performance of RBD Images (IOPS/Throughput/Latency)",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "iops"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_rbd_write_ops{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Write",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_rbd_read_ops{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "iops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "iops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 8,
+ "y": 0
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_rbd_write_bytes{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Write",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_rbd_read_bytes{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Throughput",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ns"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 9,
+ "w": 8,
+ "x": 16,
+ "y": 0
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(ceph_rbd_write_latency_sum{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval]) /\n rate(ceph_rbd_write_latency_count{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Write",
+ "refId": "A"
+ },
+ {
+ "expr": "rate(ceph_rbd_read_latency_sum{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval]) /\n rate(ceph_rbd_read_latency_count{pool=\"$pool\", image=\"$image\", cluster=~\"$cluster\", }[$__rate_interval])\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{pool}} Read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Average Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ns",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "ns",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "",
+ "multi": false,
+ "name": "pool",
+ "options": [ ],
+ "query": "label_values(ceph_rbd_read_ops{cluster=~\"$cluster\", }, pool)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "",
+ "multi": false,
+ "name": "image",
+ "options": [ ],
+ "query": "label_values(ceph_rbd_read_ops{cluster=~\"$cluster\", , pool=\"$pool\"}, image)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "RBD Details",
+ "uid": "YhCYGcuZz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/files/dashboards/rbd-overview.json b/roles/kube_prometheus_stack/files/dashboards/rbd-overview.json
new file mode 100644
index 0000000..34666c6
--- /dev/null
+++ b/roles/kube_prometheus_stack/files/dashboards/rbd-overview.json
@@ -0,0 +1,885 @@
+{
+ "__inputs": [ ],
+ "__requires": [
+ {
+ "id": "grafana",
+ "name": "Grafana",
+ "type": "grafana",
+ "version": "5.4.2"
+ },
+ {
+ "id": "graph",
+ "name": "Graph",
+ "type": "panel",
+ "version": "5.0.0"
+ },
+ {
+ "id": "prometheus",
+ "name": "Prometheus",
+ "type": "datasource",
+ "version": "5.0.0"
+ },
+ {
+ "id": "table",
+ "name": "Table",
+ "type": "panel",
+ "version": "5.0.0"
+ }
+ ],
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "-- Grafana --",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "showIn": 0,
+ "tags": [ ],
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [ ],
+ "panels": [
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "short"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 0,
+ "y": 0
+ },
+ "id": 2,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "round(sum(rate(ceph_rbd_write_ops{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "A"
+ },
+ {
+ "expr": "round(sum(rate(ceph_rbd_read_ops{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "Bps"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 8,
+ "y": 0
+ },
+ "id": 3,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "round(sum(rate(ceph_rbd_write_bytes{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Write",
+ "refId": "A"
+ },
+ {
+ "expr": "round(sum(rate(ceph_rbd_read_bytes{cluster=~\"$cluster\", }[$__rate_interval])))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Throughput",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": { },
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "fillOpacity": 8,
+ "showPoints": "never"
+ },
+ "unit": "ns"
+ }
+ },
+ "fill": 1,
+ "fillGradient": 0,
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 16,
+ "y": 0
+ },
+ "id": 4,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "sideWidth": null,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [ ],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "repeat": null,
+ "seriesOverrides": [ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "round(\n sum(rate(ceph_rbd_write_latency_sum{cluster=~\"$cluster\", }[$__rate_interval])) /\n sum(rate(ceph_rbd_write_latency_count{cluster=~\"$cluster\", }[$__rate_interval]))\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Write",
+ "refId": "A"
+ },
+ {
+ "expr": "round(\n sum(rate(ceph_rbd_read_latency_sum{cluster=~\"$cluster\", }[$__rate_interval])) /\n sum(rate(ceph_rbd_read_latency_count{cluster=~\"$cluster\", }[$__rate_interval]))\n)\n",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Read",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Average Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "timeseries",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": [ ]
+ },
+ "yaxes": [
+ {
+ "format": "ns",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": 0,
+ "show": true
+ }
+ ]
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "pool"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Pool"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "image"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Image"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "IOPS"
+ },
+ {
+ "id": "unit",
+ "value": "iops"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 0,
+ "y": 7
+ },
+ "id": 5,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n (\n sort((\n rate(ceph_rbd_write_ops{cluster=~\"$cluster\", }[$__rate_interval]) +\n on (image, pool, namespace) rate(ceph_rbd_read_ops{cluster=~\"$cluster\", }[$__rate_interval])\n ))\n )\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Highest IOPS",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ }
+ ],
+ "type": "table"
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "pool"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Pool"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "image"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Image"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Throughput"
+ },
+ {
+ "id": "unit",
+ "value": "Bps"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 8,
+ "y": 7
+ },
+ "id": 6,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n sort(\n sum(\n rate(ceph_rbd_read_bytes{cluster=~\"$cluster\", }[$__rate_interval]) +\n rate(ceph_rbd_write_bytes{cluster=~\"$cluster\", }[$__rate_interval])\n ) by (pool, image, namespace)\n )\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Highest Throughput",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ }
+ ],
+ "type": "table"
+ },
+ {
+ "columns": [ ],
+ "datasource": "${datasource}",
+ "description": "RBD per-image IO statistics are disabled by default.\n\nPlease refer to https://docs.ceph.com/en/latest/mgr/prometheus/#rbd-io-statistics for information about how to enable those optionally.",
+ "fieldConfig": {
+ "defaults": {
+ "custom": {
+ "align": "null",
+ "cellOptions": {
+ "type": "auto"
+ },
+ "filterable": true,
+ "inspect": false
+ },
+ "thresholds": {
+ "mode": "absolute",
+ "steps": [
+ {
+ "color": "green",
+ "value": null
+ },
+ {
+ "color": "red",
+ "value": 80
+ }
+ ]
+ }
+ },
+ "overrides": [
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "pool"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Pool"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "image"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Image"
+ },
+ {
+ "id": "unit",
+ "value": "short"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ },
+ {
+ "matcher": {
+ "id": "byName",
+ "options": "Value"
+ },
+ "properties": [
+ {
+ "id": "displayName",
+ "value": "Latency"
+ },
+ {
+ "id": "unit",
+ "value": "ns"
+ },
+ {
+ "id": "decimals",
+ "value": 2
+ },
+ {
+ "id": "custom.align",
+ "value": null
+ }
+ ]
+ }
+ ]
+ },
+ "gridPos": {
+ "h": 7,
+ "w": 8,
+ "x": 16,
+ "y": 7
+ },
+ "id": 7,
+ "links": [ ],
+ "options": {
+ "footer": {
+ "countRows": false,
+ "enablePagination": false,
+ "fields": "",
+ "reducer": [
+ "sum"
+ ],
+ "show": false
+ },
+ "frameIndex": 1,
+ "showHeader": true
+ },
+ "pluginVersion": "10.4.0",
+ "styles": "",
+ "targets": [
+ {
+ "expr": "topk(10,\n sum(\n rate(ceph_rbd_write_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n clamp_min(rate(ceph_rbd_write_latency_count{cluster=~\"$cluster\", }[$__rate_interval]), 1) +\n rate(ceph_rbd_read_latency_sum{cluster=~\"$cluster\", }[$__rate_interval]) /\n clamp_min(rate(ceph_rbd_read_latency_count{cluster=~\"$cluster\", }[$__rate_interval]), 1)\n ) by (pool, image, namespace)\n)\n",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "A"
+ }
+ ],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Highest Latency",
+ "transformations": [
+ {
+ "id": "merge",
+ "options": {
+ "reducers": [ ]
+ }
+ }
+ ],
+ "type": "table"
+ }
+ ],
+ "refresh": "30s",
+ "rows": [ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph-mixin",
+ "overview"
+ ],
+ "templating": {
+ "list": [
+ {
+ "current": {
+ "text": "default",
+ "value": "default"
+ },
+ "hide": 0,
+ "label": "Data Source",
+ "name": "datasource",
+ "options": [ ],
+ "query": "prometheus",
+ "refresh": 1,
+ "regex": "",
+ "type": "datasource"
+ },
+ {
+ "allValue": null,
+ "current": { },
+ "datasource": "$datasource",
+ "hide": 0,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [ ],
+ "query": "label_values(ceph_health_status, cluster)",
+ "refresh": 1,
+ "regex": "(.*)",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "",
+ "title": "RBD Overview",
+ "uid": "41FrpeUiz",
+ "version": 0
+}
diff --git a/roles/kube_prometheus_stack/tasks/main.yml b/roles/kube_prometheus_stack/tasks/main.yml
index 1d107b0..2b56430 100644
--- a/roles/kube_prometheus_stack/tasks/main.yml
+++ b/roles/kube_prometheus_stack/tasks/main.yml
@@ -313,6 +313,26 @@
state: present
- name: node-exporter-full
state: present
+ - name: ceph-cluster
+ state: present
+ - name: ceph-cluster-advanced
+ state: present
+ - name: hosts-overview
+ state: present
+ - name: host-details
+ state: present
+ - name: pool-overview
+ state: present
+ - name: pool-detail
+ state: present
+ - name: osds-overview
+ state: present
+ - name: osd-device-details
+ state: present
+ - name: rbd-overview
+ state: present
+ - name: rbd-details
+ state: present
tags:
- kube-prometheus-stack-dashboards