Merge "Added parameter for enabling legacy monitoring in newer ceph versions. Related-Prod: PROD-35190"
diff --git a/ceph/files/grafana_dashboards/ceph-cluster_prometheus.json b/ceph/files/grafana_dashboards/luminous-ceph-cluster_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/ceph-cluster_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-ceph-cluster_prometheus.json
diff --git a/ceph/files/grafana_dashboards/hosts-overview_prometheus.json b/ceph/files/grafana_dashboards/luminous-hosts-overview_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/hosts-overview_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-hosts-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/osds-detail_prometheus.json b/ceph/files/grafana_dashboards/luminous-osds-detail_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/osds-detail_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-osds-detail_prometheus.json
diff --git a/ceph/files/grafana_dashboards/osds-overview_prometheus.json b/ceph/files/grafana_dashboards/luminous-osds-overview_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/osds-overview_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-osds-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/pool-overview_prometheus.json b/ceph/files/grafana_dashboards/luminous-pool-overview_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/pool-overview_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-pool-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/radosgw-detail_prometheus.json b/ceph/files/grafana_dashboards/luminous-radosgw-detail_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/radosgw-detail_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-radosgw-detail_prometheus.json
diff --git a/ceph/files/grafana_dashboards/radosgw-overview_prometheus.json b/ceph/files/grafana_dashboards/luminous-radosgw-overview_prometheus.json
similarity index 100%
rename from ceph/files/grafana_dashboards/radosgw-overview_prometheus.json
rename to ceph/files/grafana_dashboards/luminous-radosgw-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/nautilus-ceph-cluster_prometheus.json b/ceph/files/grafana_dashboards/nautilus-ceph-cluster_prometheus.json
new file mode 100644
index 0000000..dedf7e9
--- /dev/null
+++ b/ceph/files/grafana_dashboards/nautilus-ceph-cluster_prometheus.json
@@ -0,0 +1,1300 @@
+{%- raw %}
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "prometheus",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "description": "Ceph cluster overview",
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "id": null,
+ "iteration": 1545072771425,
+ "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": null,
+ "editable": true,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "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{instance=~'$instance'}",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "thresholds": "1,2",
+ "timeFrom": "1m",
+ "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"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 128, 45, 0.9)",
+ "rgba(237, 129, 40, 0.9)",
+ "rgb(255, 0, 0)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 2,
+ "y": 0
+ },
+ "hideTimeOverride": true,
+ "id": 67,
+ "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": "sum(ceph_osd_stat_bytes{instance=~\"$instance\"})",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": "1m",
+ "title": "Total",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "#299c46",
+ "rgba(237, 129, 40, 0.89)",
+ "#d44a3a"
+ ],
+ "datasource": null,
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 6,
+ "w": 4,
+ "x": 4,
+ "y": 0
+ },
+ "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{instance=~\"$instance\"})/sum(ceph_osd_stat_bytes{instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Used",
+ "refId": "A"
+ }
+ ],
+ "thresholds": "70,80",
+ "title": "Capacity used",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 8,
+ "x": 8,
+ "y": 0
+ },
+ "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": "ceph_pg_total",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Total",
+ "refId": "A"
+ },
+ {
+ "expr": "ceph_pg_active",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Active",
+ "refId": "B"
+ },
+ {
+ "expr": "ceph_pg_total - ceph_pg_active",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Inactive",
+ "refId": "G"
+ },
+ {
+ "expr": "ceph_pg_undersized",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Undersized",
+ "refId": "F"
+ },
+ {
+ "expr": "ceph_pg_degraded",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Degraded",
+ "refId": "C"
+ },
+ {
+ "expr": "ceph_pg_inconsistent",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Inconsistent",
+ "refId": "D"
+ },
+ {
+ "expr": "ceph_pg_down",
+ "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": "graph",
+ "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
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 0,
+ "gridPos": {
+ "h": 6,
+ "w": 8,
+ "x": 16,
+ "y": 0
+ },
+ "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{instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Apply Latency P_95",
+ "refId": "A"
+ },
+ {
+ "expr": "quantile(0.95, ceph_osd_commit_latency_ms{instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Commit Latency P_95",
+ "refId": "B"
+ },
+ {
+ "expr": "avg(ceph_osd_apply_latency_ms{instance=~\"$instance\"})",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Avg Apply Latency",
+ "refId": "C"
+ },
+ {
+ "expr": "avg(ceph_osd_commit_latency_ms{instance=~\"$instance\"})",
+ "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": "graph",
+ "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
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 128, 45, 0.9)",
+ "rgba(237, 129, 40, 0.9)",
+ "rgb(255, 0, 0)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 0,
+ "y": 3
+ },
+ "hideTimeOverride": true,
+ "id": 68,
+ "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": "sum(ceph_osd_stat_bytes{instance=~\"$instance\"})-sum(ceph_osd_stat_bytes_used{instance=~\"$instance\"})",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": "1m",
+ "title": "Free",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [],
+ "valueName": "current"
+ },
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 128, 45, 0.9)",
+ "rgba(237, 129, 40, 0.9)",
+ "rgb(255, 0, 0)"
+ ],
+ "datasource": null,
+ "editable": true,
+ "error": false,
+ "format": "bytes",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "gridPos": {
+ "h": 3,
+ "w": 2,
+ "x": 2,
+ "y": 3
+ },
+ "hideTimeOverride": true,
+ "id": 69,
+ "interval": "1m",
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 2,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(ceph_osd_stat_bytes_used{instance=~\"$instance\"})",
+ "format": "time_series",
+ "interval": "$interval",
+ "intervalFactor": 1,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "thresholds": "",
+ "timeFrom": "1m",
+ "title": "Used",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "50%",
+ "valueMaps": [],
+ "valueName": "current"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 0,
+ "y": 6
+ },
+ "id": 45,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 0.5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "Reads",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(irate(ceph_osd_op_w_in_bytes{instance=~\"$instance\"}[$rate_interval]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "A"
+ },
+ {
+ "expr": "sum(irate(ceph_osd_op_r_out_bytes{instance=~\"$instance\"}[$rate_interval]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Cluster I/O",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 12,
+ "x": 12,
+ "y": 6
+ },
+ "id": 62,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(deriv(ceph_pool_stored{instance=~\"$instance\"}[$rate_interval]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Bytes",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "In-/Egress",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": " Egress (-) / Ingress (+)",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": 1
+ },
+ "color": {
+ "cardColor": "rgb(0, 254, 255)",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateBlues",
+ "exponent": 0.5,
+ "min": null,
+ "mode": "spectrum"
+ },
+ "dataFormat": "timeseries",
+ "datasource": null,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 15
+ },
+ "heatmap": {},
+ "highlightCards": true,
+ "id": 55,
+ "legend": {
+ "show": true
+ },
+ "links": [],
+ "span": 12,
+ "targets": [
+ {
+ "expr": "ceph_osd_stat_bytes_used{instance=~'$instance'} / ceph_osd_stat_bytes{instance=~'$instance'}",
+ "format": "time_series",
+ "instant": false,
+ "interval": "1m",
+ "intervalFactor": 1,
+ "legendFormat": "Util (%)",
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "timeFrom": null,
+ "title": "OSD Capacity Utilization",
+ "tooltip": {
+ "show": true,
+ "showHistogram": false
+ },
+ "type": "heatmap",
+ "xAxis": {
+ "show": true
+ },
+ "xBucketNumber": null,
+ "xBucketSize": "",
+ "yAxis": {
+ "decimals": null,
+ "format": "percentunit",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true,
+ "splitFactor": null
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "cards": {
+ "cardPadding": null,
+ "cardRound": 1
+ },
+ "color": {
+ "cardColor": "#b4ff00",
+ "colorScale": "sqrt",
+ "colorScheme": "interpolateBlues",
+ "exponent": 0.5,
+ "mode": "spectrum"
+ },
+ "dataFormat": "timeseries",
+ "datasource": null,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 15
+ },
+ "heatmap": {},
+ "highlightCards": true,
+ "id": 59,
+ "legend": {
+ "show": true
+ },
+ "links": [],
+ "targets": [
+ {
+ "expr": "ceph_osd_numpg{instance=~\"$instance\"}",
+ "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
+ },
+ "yBucketBound": "auto",
+ "yBucketNumber": null,
+ "yBucketSize": null
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "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[$rate_interval]))",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Op/s",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Recovery Rate",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": "Recovery Ops/s",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "refresh": "30s",
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph",
+ "cluster"
+ ],
+ "templating": {
+ "list": [
+ {
+ "auto": true,
+ "auto_count": 10,
+ "auto_min": "1m",
+ "current": {
+ "text": "auto",
+ "value": "$__auto_interval_interval"
+ },
+ "datasource": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "Interval",
+ "multi": false,
+ "name": "interval",
+ "options": [
+ {
+ "selected": true,
+ "text": "auto",
+ "value": "$__auto_interval_interval"
+ },
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "30m",
+ "value": "30m"
+ },
+ {
+ "selected": false,
+ "text": "1h",
+ "value": "1h"
+ },
+ {
+ "selected": false,
+ "text": "6h",
+ "value": "6h"
+ },
+ {
+ "selected": false,
+ "text": "12h",
+ "value": "12h"
+ },
+ {
+ "selected": false,
+ "text": "1d",
+ "value": "1d"
+ },
+ {
+ "selected": false,
+ "text": "7d",
+ "value": "7d"
+ },
+ {
+ "selected": false,
+ "text": "14d",
+ "value": "14d"
+ },
+ {
+ "selected": false,
+ "text": "30d",
+ "value": "30d"
+ }
+ ],
+ "query": "1m,10m,30m,1h,6h,12h,1d,7d,14d,30d",
+ "refresh": 2,
+ "skipUrlSync": false,
+ "type": "interval"
+ },
+ {
+ "allFormat": "glob",
+ "allValue": null,
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": null,
+ "hide": 0,
+ "hideLabel": false,
+ "includeAll": true,
+ "label": "Exporter Instance",
+ "multi": false,
+ "multiFormat": "glob",
+ "name": "instance",
+ "options": [],
+ "query": "label_values(ceph_health_status, instance)",
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "tags": [],
+ "text": "3m",
+ "value": "3m"
+ },
+ "hide": 0,
+ "includeAll": false,
+ "label": "rate_interval",
+ "multi": false,
+ "name": "rate_interval",
+ "options": [
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": true,
+ "text": "3m",
+ "value": "3m"
+ },
+ {
+ "selected": false,
+ "text": "5m",
+ "value": "5m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "15m",
+ "value": "15m"
+ }
+ ],
+ "query": "1m,3m,5m,10m,15m",
+ "skipUrlSync": false,
+ "type": "custom"
+ }
+ ]
+ },
+ "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": "browser",
+ "title": "Ceph - Cluster",
+ "version": 7
+}
+{%- endraw %}
diff --git a/ceph/files/grafana_dashboards/hosts-overview_prometheus.json b/ceph/files/grafana_dashboards/nautilus-hosts-overview_prometheus.json
similarity index 100%
copy from ceph/files/grafana_dashboards/hosts-overview_prometheus.json
copy to ceph/files/grafana_dashboards/nautilus-hosts-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/nautilus-osds-detail_prometheus.json b/ceph/files/grafana_dashboards/nautilus-osds-detail_prometheus.json
new file mode 100644
index 0000000..cc2b4d1
--- /dev/null
+++ b/ceph/files/grafana_dashboards/nautilus-osds-detail_prometheus.json
@@ -0,0 +1,789 @@
+{%- raw %}
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "prometheus",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "id": null,
+ "iteration": 1545072619802,
+ "links": [],
+ "panels": [
+ {
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 14,
+ "panels": [],
+ "title": "OSD Performance",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 1
+ },
+ "id": 2,
+ "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": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(ceph_osd_op_r_latency_sum{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[$rate_interval]) * 1000",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "READs",
+ "refId": "A"
+ },
+ {
+ "expr": "irate(ceph_osd_op_w_latency_sum{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[$rate_interval]) * 1000",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "WRITEs",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD $osd_id Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 1
+ },
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(ceph_osd_op_r{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "irate(ceph_osd_op_w{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD $osd_id R/W IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 12,
+ "y": 1
+ },
+ "id": 7,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "irate(ceph_osd_op_r_out_bytes{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Read Bytes",
+ "refId": "A"
+ },
+ {
+ "expr": "irate(ceph_osd_op_w_in_bytes{ceph_daemon=~\"osd.[[osd_id]]\"}[$rate_interval])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "Write Bytes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "OSD $osd_id R/W Bytes",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 10
+ },
+ "id": 12,
+ "panels": [],
+ "title": "Physical Device Performance",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 0,
+ "y": 11
+ },
+ "id": 9,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(label_replace(irate(diskio_read_time[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") / label_replace(label_replace(irate(diskio_reads[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}/{{device}} Reads",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(label_replace(irate(diskio_write_time[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") / label_replace(label_replace(irate(diskio_writes[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}}/{{device}} Writes",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device Latency for OSD $osd_id",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ms",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 6,
+ "y": 11
+ },
+ "id": 5,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(label_replace(irate(diskio_reads[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} READS",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(label_replace(irate(diskio_writes[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} WRITES",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device R/W IOPS for OSD $osd_id",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 12,
+ "y": 11
+ },
+ "id": 10,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(label_replace(irate(diskio_read_bytes[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} READS",
+ "refId": "A"
+ },
+ {
+ "expr": "label_replace(label_replace(irate(diskio_write_bytes[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.+)\"), \"device\", \"$1\", \"name\", \"(.+)\") and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}} WRITES",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device R/W Bytes for OSD $osd_id",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 6,
+ "x": 18,
+ "y": 11
+ },
+ "id": 4,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "label_replace(irate(diskio_io_time[$rate_interval]), \"instance\", \"$1\", \"host\", \"(.*)\") / 1000 and on (name,instance) label_replace(label_replace(ceph_disk_occupation{ceph_daemon=~\"osd.$osd_id\"}, \"name\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1\", \"instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{instance}} {{device}}",
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Physical Device Util% for OSD $osd_id",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "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
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ }
+ ],
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph",
+ "osd"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+ "text": "0",
+ "value": "0"
+ },
+ "datasource": null,
+ "hide": 0,
+ "includeAll": false,
+ "label": "OSD Id",
+ "multi": false,
+ "name": "osd_id",
+ "options": [],
+ "query": "label_values(ceph_osd_metadata,ceph_daemon)",
+ "refresh": 1,
+ "regex": "osd.(.*)",
+ "skipUrlSync": false,
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "3m",
+ "value": "3m"
+ },
+ "hide": 0,
+ "includeAll": false,
+ "label": "rate_interval",
+ "multi": false,
+ "name": "rate_interval",
+ "options": [
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": true,
+ "text": "3m",
+ "value": "3m"
+ },
+ {
+ "selected": false,
+ "text": "5m",
+ "value": "5m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "15m",
+ "value": "15m"
+ }
+ ],
+ "query": "1m,3m,5m,10m,15m",
+ "skipUrlSync": false,
+ "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 OSD device details",
+ "version": 8
+}
+{%- endraw %}
diff --git a/ceph/files/grafana_dashboards/osds-overview_prometheus.json b/ceph/files/grafana_dashboards/nautilus-osds-overview_prometheus.json
similarity index 100%
copy from ceph/files/grafana_dashboards/osds-overview_prometheus.json
copy to ceph/files/grafana_dashboards/nautilus-osds-overview_prometheus.json
diff --git a/ceph/files/grafana_dashboards/nautilus-pool-overview_prometheus.json b/ceph/files/grafana_dashboards/nautilus-pool-overview_prometheus.json
new file mode 100644
index 0000000..39e94af
--- /dev/null
+++ b/ceph/files/grafana_dashboards/nautilus-pool-overview_prometheus.json
@@ -0,0 +1,1245 @@
+{%- raw %}
+{
+ "annotations": {
+ "list": [
+ {
+ "builtIn": 1,
+ "datasource": "prometheus",
+ "enable": true,
+ "hide": true,
+ "iconColor": "rgba(0, 211, 255, 1)",
+ "name": "Annotations & Alerts",
+ "type": "dashboard"
+ }
+ ]
+ },
+ "editable": true,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "iteration": 1574692480292,
+ "links": [],
+ "panels": [
+ {
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 0
+ },
+ "id": 14,
+ "panels": [],
+ "repeat": null,
+ "title": "Pool Overview",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 1
+ },
+ "id": 1,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(rate(ceph_pool_rd{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id,instance) group_left(name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance)",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "F"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Read IOPS by Pool",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "rps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 1
+ },
+ "id": 16,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(rate(ceph_pool_wr{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id,instance) group_left(name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance)",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "F"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Write IOPS by Pool",
+ "tooltip": {
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "wps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 8
+ },
+ "id": 2,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(rate(ceph_pool_rd_bytes{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id,instance) group_left(name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Read Throughput by Pool",
+ "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": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": null,
+ "fill": 2,
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 8
+ },
+ "id": 17,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "avg(rate(ceph_pool_wr_bytes{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id,instance) group_left(name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance)",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{name}}",
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Write Throughput by Pool",
+ "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": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
+ },
+ {
+ "collapsed": false,
+ "gridPos": {
+ "h": 1,
+ "w": 24,
+ "x": 0,
+ "y": 15
+ },
+ "id": 15,
+ "panels": [],
+ "repeat": null,
+ "title": "Top 5's",
+ "type": "row"
+ },
+ {
+ "columns": [
+ {
+ "text": "Current",
+ "value": "current"
+ }
+ ],
+ "datasource": null,
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 16
+ },
+ "id": 3,
+ "links": [],
+ "minSpan": 12,
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 7,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Time",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "id",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "instance",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "job",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool Name",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "name",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool ID",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "pool_id",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Read IOPS",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "pattern": "Value",
+ "thresholds": [],
+ "type": "number",
+ "unit": "rps"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "topk(5,avg(rate(ceph_pool_rd{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance))",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "title": "Top 5 Pools by Read IOPS",
+ "transform": "table",
+ "type": "table"
+ },
+ {
+ "columns": [
+ {
+ "text": "Current",
+ "value": "current"
+ }
+ ],
+ "datasource": null,
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 16
+ },
+ "id": 20,
+ "links": [],
+ "minSpan": 12,
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 7,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Time",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "id",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "instance",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "job",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool Name",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "name",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool ID",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "pool_id",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Write IOPS",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "pattern": "Value",
+ "thresholds": [],
+ "type": "number",
+ "unit": "wps"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "topk(5,avg(rate(ceph_pool_wr{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance))",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "title": "Top 5 Pools by Write IOPS",
+ "transform": "table",
+ "type": "table"
+ },
+ {
+ "columns": [
+ {
+ "text": "Current",
+ "value": "current"
+ }
+ ],
+ "datasource": null,
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 23
+ },
+ "id": 21,
+ "links": [],
+ "minSpan": 12,
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 7,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Time",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "id",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "instance",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "job",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool Name",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "name",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool ID",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "pool_id",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Read Throughput",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "pattern": "Value",
+ "thresholds": [],
+ "type": "number",
+ "unit": "Bps"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "topk(5,avg(rate(ceph_pool_rd_bytes{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance))",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "title": "Top 5 Pools by Read Throughput",
+ "transform": "table",
+ "type": "table"
+ },
+ {
+ "columns": [
+ {
+ "text": "Current",
+ "value": "current"
+ }
+ ],
+ "datasource": null,
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 12,
+ "y": 23
+ },
+ "id": 22,
+ "links": [],
+ "minSpan": 12,
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 7,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Time",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "id",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "instance",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "job",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool Name",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "name",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool ID",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "pool_id",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Write Throughput",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 0,
+ "pattern": "Value",
+ "thresholds": [],
+ "type": "number",
+ "unit": "Bps"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "topk(5,avg(rate(ceph_pool_wr_bytes{pool_id=~\"[[pool_id]]\"}[$rate_interval]) * on(pool_id) group_left(instance,name) ceph_pool_metadata{pool_id=~\"[[pool_id]]\"}) without (instance))",
+ "format": "table",
+ "instant": true,
+ "intervalFactor": 2,
+ "refId": "A",
+ "textEditor": true
+ }
+ ],
+ "title": "Top 5 Pools by Write Throughput",
+ "transform": "table",
+ "type": "table"
+ },
+ {
+ "columns": [],
+ "datasource": null,
+ "fontSize": "100%",
+ "gridPos": {
+ "h": 7,
+ "w": 12,
+ "x": 0,
+ "y": 30
+ },
+ "id": 5,
+ "links": [],
+ "minSpan": 8,
+ "pageSize": null,
+ "scroll": true,
+ "showHeader": true,
+ "sort": {
+ "col": 6,
+ "desc": true
+ },
+ "styles": [
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Time",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "instance",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "job",
+ "thresholds": [],
+ "type": "hidden",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool Name",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "name",
+ "thresholds": [],
+ "type": "string",
+ "unit": "short"
+ },
+ {
+ "alias": "Pool ID",
+ "colorMode": null,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "pool_id",
+ "thresholds": [],
+ "type": "number",
+ "unit": "short"
+ },
+ {
+ "alias": "Capacity Used",
+ "colorMode": "value",
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "dateFormat": "YYYY-MM-DD HH:mm:ss",
+ "decimals": 2,
+ "pattern": "Value",
+ "thresholds": [
+ "70",
+ "85"
+ ],
+ "type": "number",
+ "unit": "percentunit"
+ }
+ ],
+ "targets": [
+ {
+ "expr": "topk(5,avg((ceph_pool_stored / (ceph_pool_stored + ceph_pool_max_avail)) * on(pool_id) group_left(name) ceph_pool_metadata) without (instance))",
+ "format": "table",
+ "hide": false,
+ "instant": true,
+ "intervalFactor": 1,
+ "legendFormat": "",
+ "refId": "D"
+ }
+ ],
+ "title": "Top 5 Pools By Capacity Used",
+ "transform": "table",
+ "type": "table"
+ }
+ ],
+ "refresh": "15s",
+ "schemaVersion": 16,
+ "style": "dark",
+ "tags": [
+ "ceph",
+ "pool"
+ ],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": null,
+ "hide": 2,
+ "includeAll": true,
+ "label": null,
+ "multi": false,
+ "name": "pool_id",
+ "options": [],
+ "query": "label_values(ceph_pool_metadata,pool_id)",
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "text": "All",
+ "value": "$__all"
+ },
+ "datasource": null,
+ "hide": 2,
+ "includeAll": true,
+ "label": "Pool Name",
+ "multi": false,
+ "name": "pool_name",
+ "options": [],
+ "query": "label_values(ceph_pool_metadata,name)",
+ "refresh": 1,
+ "regex": "",
+ "skipUrlSync": false,
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+ "tags": [],
+ "text": "3m",
+ "value": "3m"
+ },
+ "hide": 0,
+ "includeAll": false,
+ "label": "rate_interval",
+ "multi": false,
+ "name": "rate_interval",
+ "options": [
+ {
+ "selected": false,
+ "text": "1m",
+ "value": "1m"
+ },
+ {
+ "selected": true,
+ "text": "3m",
+ "value": "3m"
+ },
+ {
+ "selected": false,
+ "text": "5m",
+ "value": "5m"
+ },
+ {
+ "selected": false,
+ "text": "10m",
+ "value": "10m"
+ },
+ {
+ "selected": false,
+ "text": "15m",
+ "value": "15m"
+ }
+ ],
+ "query": "1m,3m,5m,10m,15m",
+ "skipUrlSync": false,
+ "type": "custom"
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "15s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Ceph Pools Overview",
+ "version": 12
+}
+{%- endraw %}
diff --git a/ceph/files/grafana_dashboards/radosgw-detail_prometheus.json b/ceph/files/grafana_dashboards/nautilus-radosgw-detail_prometheus.json
similarity index 100%
copy from ceph/files/grafana_dashboards/radosgw-detail_prometheus.json
copy to ceph/files/grafana_dashboards/nautilus-radosgw-detail_prometheus.json
diff --git a/ceph/files/grafana_dashboards/radosgw-overview_prometheus.json b/ceph/files/grafana_dashboards/nautilus-radosgw-overview_prometheus.json
similarity index 100%
copy from ceph/files/grafana_dashboards/radosgw-overview_prometheus.json
copy to ceph/files/grafana_dashboards/nautilus-radosgw-overview_prometheus.json
diff --git a/ceph/meta/grafana.yml b/ceph/meta/grafana.yml
index a475784..7c05f3e 100644
--- a/ceph/meta/grafana.yml
+++ b/ceph/meta/grafana.yml
@@ -19,30 +19,30 @@
ceph_cluster_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/ceph-cluster_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-ceph-cluster_prometheus.json
ceph_osd_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/osds-overview_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-osds-overview_prometheus.json
ceph_osd_details_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/osds-detail_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-osds-detail_prometheus.json
ceph_pools_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/pool-overview_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-pool-overview_prometheus.json
ceph_hosts_overview_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/hosts-overview_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-hosts-overview_prometheus.json
ceph_radosgw_overview_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/radosgw-overview_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-radosgw-overview_prometheus.json
ceph_radosgw_detail_prometheus:
datasource: prometheus
format: json
- template: ceph/files/grafana_dashboards/radosgw-detail_prometheus.json
+ template: ceph/files/grafana_dashboards/{{ common.version }}-radosgw-detail_prometheus.json
{%- endif %}
{%- endif %}
diff --git a/ceph/meta/prometheus.yml b/ceph/meta/prometheus.yml
index 77558b2..01a6322 100644
--- a/ceph/meta/prometheus.yml
+++ b/ceph/meta/prometheus.yml
@@ -156,6 +156,7 @@
{%- else -%}
{%- if mon is defined and mon.get('enabled') %}
+ {%- set pool_used_space = 'ceph_pool_bytes_used' if common.version in ['luminous'] else 'ceph_pool_stored' -%}
{%- raw %}
server:
recording:
@@ -366,7 +367,7 @@
CephPool{{pool_name|replace(".", "")|replace("-", "")}}SpaceUsageWarning:
{%- set threshold = monitoring_pool.pool_space_used_utilization_warning_threshold|default('0.75')|float %}
if: >-
- ceph_pool_bytes_used / (ceph_pool_bytes_used + ceph_pool_max_avail) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > {{threshold}}
+ {{pool_used_space}} / ({{pool_used_space}} + ceph_pool_max_avail) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > {{threshold}}
for: 3m
labels:
severity: warning
@@ -377,7 +378,7 @@
CephPool{{pool_name|replace(".", "")|replace("-", "")}}SpaceUsageCritical:
{%- set threshold = monitoring_pool.pool_space_used_critical_threshold|default('0.85')|float %}
if: >-
- ceph_pool_bytes_used / (ceph_pool_bytes_used + ceph_pool_max_avail) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > {{threshold}}
+ {{pool_used_space}} / ({{pool_used_space}} + ceph_pool_max_avail) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > {{threshold}}
for: 3m
labels:
severity: minor
@@ -390,7 +391,7 @@
{%- set threshold = monitoring.prediction_threshold %}
{%- set space_threshold = monitoring_pool.space_threshold %}
if: >-
- predict_linear(ceph_pool_bytes_used[{{threshold}}d], {{threshold}} * 86400) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > (ceph_pool_bytes_used + ceph_pool_max_avail) * {{space_threshold}} * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"}
+ predict_linear({{pool_used_space}}[{{threshold}}d], {{threshold}} * 86400) * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"} > ({{pool_used_space}} + ceph_pool_max_avail) * {{space_threshold}} * on(pool_id) group_left(name) ceph_pool_metadata{name="{{pool_name}}"}
for: 30m
labels:
severity: minor
diff --git a/ceph/setup/keyring.sls b/ceph/setup/keyring.sls
index f9fbbf8..d718047 100644
--- a/ceph/setup/keyring.sls
+++ b/ceph/setup/keyring.sls
@@ -56,9 +56,9 @@
{%- if client['entity'] == "client." + keyring_client_name %}
{%- for cap_name, cap in client.caps.iteritems() %}
{%- if cap != keyring.caps[cap_name] %}
-ceph_update_caps_for_{{ keyring_client_name }}:
+ceph_update_caps_{{ cap_name }}_for_{{ keyring_client_name }}:
cmd.run:
- - name: ceph -c /etc/ceph/{{ ceph_cluster }}.conf auth caps client.{{ keyring_client_name }} {%- for cap_name, cap in keyring.caps.iteritems() %} {{ cap_name }} '{{ cap }}' {%- endfor %}
+ - name: ceph -c /etc/ceph/{{ ceph_cluster }}.conf auth caps client.{{ keyring_client_name }} {{ cap_name }} '{{ cap }}'
{%- endif %}
{%- endfor %}
{%- endif %}