{
+ "__inputs": [],
"__requires": [
{
"type": "grafana",
"id": "grafana",
"name": "Grafana",
- "version": "5.0.0"
+ "version": "5.3.2"
},
{
"type": "panel",
"gnetId": null,
"graphTooltip": 0,
"id": null,
- "iteration": 1538021829737,
+ "iteration": 1557386759572,
"links": [
{
"asDropdown": true,
"dashes": false,
"datasource": "$datasource",
"description": "Shows the CPU breakdown. When multiple servers are selected, only the first host's cpu data is shown",
- "fill": 3,
+ "fill": 1,
"gridPos": {
"h": 10,
"w": 6,
},
"yaxes": [
{
- "format": "short",
- "label": "",
+ "format": "percent",
+ "label": "% Utilization",
"logBase": 1,
"max": "100",
- "min": "0",
+ "min": null,
"show": true
},
{
"min": null,
"show": false
}
- ]
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
"aliasColors": {
"yaxes": [
{
"format": "bytes",
- "label": null,
+ "label": "RAM used",
"logBase": 1,
"max": null,
"min": "0",
"min": null,
"show": false
}
- ]
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
"aliasColors": {},
"dashes": false,
"datasource": "$datasource",
"description": "Show the network load (rx,tx) across all interfaces (excluding loopback 'lo')",
- "fill": 1,
+ "fill": 0,
"gridPos": {
"h": 10,
"w": 6,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "seriesOverrides": [],
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
"spaceLength": 10,
- "stack": true,
+ "stack": false,
"steppedLine": false,
"targets": [
{
"yaxes": [
{
"format": "decbytes",
- "label": null,
+ "label": "Send (-) / Receive (+)",
"logBase": 1,
"max": null,
- "min": "0",
+ "min": null,
"show": true
},
{
"min": null,
"show": false
}
- ]
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
"aliasColors": {},
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
+ "fill": 0,
"gridPos": {
"h": 5,
"w": 3,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "seriesOverrides": [],
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
"spaceLength": 10,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "irate(node_network_transmit_drop{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_transmit_drop_total{instance=~\"[[ceph_hosts]].*\"}[1m]) + \nirate(node_network_receive_drop{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_receive_drop_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
+ "expr": "irate(node_network_receive_drop{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_receive_drop_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
"format": "time_series",
"instant": false,
"intervalFactor": 1,
- "legendFormat": "{{device}}",
+ "legendFormat": "{{device}}.rx",
"refId": "A"
+ },
+ {
+ "expr": "irate(node_network_transmit_drop{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_transmit_drop_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.tx",
+ "refId": "B"
}
],
"thresholds": [],
"timeFrom": "15m",
"timeShift": null,
- "title": "Network Drops (last 15mins)",
+ "title": "Network drop rate",
"tooltip": {
"shared": true,
"sort": 0,
},
"yaxes": [
{
- "format": "short",
- "label": null,
+ "format": "pps",
+ "label": "Send (-) / Receive (+)",
"logBase": 1,
"max": null,
"min": "0",
"min": null,
"show": false
}
- ]
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
"cacheTimeout": null,
"tableColumn": "",
"targets": [
{
- "expr": "sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{instance=~\"($ceph_hosts).*\"})",
+ "expr": "sum(ceph_osd_stat_bytes and on (ceph_daemon) ceph_disk_occupation{exported_instance=~\"($ceph_hosts).*\"})",
"format": "time_series",
"intervalFactor": 2,
"refId": "A",
"dashLength": 10,
"dashes": false,
"datasource": "$datasource",
- "fill": 1,
+ "fill": 0,
"gridPos": {
"h": 5,
"w": 3,
"pointradius": 5,
"points": false,
"renderer": "flot",
- "seriesOverrides": [],
+ "seriesOverrides": [
+ {
+ "alias": "/.*tx/",
+ "transform": "negative-Y"
+ }
+ ],
"spaceLength": 10,
"stack": false,
"steppedLine": false,
"targets": [
{
- "expr": "irate(node_network_transmit_errs{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_transmit_errs_total{instance=~\"[[ceph_hosts]].*\"}[1m])+ \nirate(node_network_receive_errs{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_receive_errs_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
+ "expr": "irate(node_network_receive_errs{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_receive_errs_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
"format": "time_series",
"instant": false,
"intervalFactor": 1,
- "legendFormat": "{{device}}",
+ "legendFormat": "{{device}}.rx",
"refId": "A"
+ },
+ {
+ "expr": "irate(node_network_transmit_errs{instance=~\"[[ceph_hosts]].*\"}[1m]) or irate(node_network_transmit_errs_total{instance=~\"[[ceph_hosts]].*\"}[1m])",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}.tx",
+ "refId": "B"
}
],
"thresholds": [],
"timeFrom": "15m",
"timeShift": null,
- "title": "Network Errors(last 15mins)",
+ "title": "Network error rate",
"tooltip": {
"shared": true,
"sort": 0,
},
"yaxes": [
{
- "format": "short",
- "label": null,
+ "format": "pps",
+ "label": "Send (-) / Receive (+)",
"logBase": 1,
"max": null,
"min": "0",
"min": null,
"show": false
}
- ]
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
},
{
- "collapsed": true,
+ "collapsed": false,
"gridPos": {
"h": 1,
"w": 24,
"y": 11
},
"id": 12,
- "panels": [
+ "panels": [],
+ "repeat": null,
+ "title": "OSD Disk Performance Statistics",
+ "type": "row"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "$datasource",
+ "description": "For any OSD devices on the host, this chart shows the iops per physical device. Each device is shown by it's name and corresponding OSD id value",
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 0,
+ "y": 12
+ },
+ "id": 6,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$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",
- "fill": 1,
- "gridPos": {
- "h": 9,
- "w": 11,
- "x": 0,
- "y": 12
- },
- "id": 6,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "minSpan": 12,
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "((irate(node_disk_reads_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts).*\"}[5m]) ) + \n(irate(node_disk_writes_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m]))) *\non(instance, device) group_left(osd_id) label_replace(ceph_disk_occupation, \"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\") ",
- "format": "time_series",
- "intervalFactor": 1,
- "legendFormat": "{{device}}({{osd_id}})",
- "refId": "A",
- "step": 10,
- "textEditor": true
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "$ceph_hosts Disk IOPS",
- "tooltip": {
- "shared": true,
- "sort": 2,
- "value_type": "individual"
- },
- "type": "graph",
- "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": null,
- "show": false
- }
- ]
+ "alias": "/.*reads/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "(irate(node_disk_writes_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) writes",
+ "refId": "A",
+ "step": 10,
+ "textEditor": true
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "description": "Show disk utilization % (util) of any OSD devices on the host by the physical device name and associated OSD id.",
- "fill": 1,
- "gridPos": {
- "h": 9,
- "w": 11,
- "x": 12,
- "y": 12
- },
- "id": 5,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "minSpan": 12,
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "(\n (irate(node_disk_io_time_ms[5m]) / 10 ) or\n irate(node_disk_io_time_seconds_total[5m]) * 100)\n* on(instance, device) group_left(osd_id) label_replace(ceph_disk_occupation{instance=~\"($ceph_hosts).*\"}, \"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")\n",
- "format": "time_series",
- "hide": false,
- "intervalFactor": 1,
- "legendFormat": "{{device}}({{osd_id}})",
- "refId": "C",
- "step": 10,
- "textEditor": true
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "$ceph_hosts Disk utilisation",
- "tooltip": {
- "shared": true,
- "sort": 2,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "%Util",
- "logBase": 1,
- "max": "100",
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
+ "expr": "(irate(node_disk_reads_completed{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_reads_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) reads",
+ "refId": "B"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk IOPS",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": "Read (-) / Write (+)",
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "show": true
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$datasource",
- "description": "For OSD hosts, this chart shows the latency at the physical drive. Each drive is shown by device name, with it's corresponding OSD id",
- "fill": 1,
- "gridPos": {
- "h": 9,
- "w": 11,
- "x": 0,
- "y": 21
- },
- "id": 7,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "minSpan": 12,
- "nullPointMode": "null as zero",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "max by(instance,device) (((\n (irate(node_disk_write_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_writes_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n (irate(node_disk_read_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_reads_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n ) / 1000) or\n (irate(node_disk_write_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_writes_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n (irate(node_disk_read_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_reads_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n ) *\n on(instance,device) group_left(osd_id) label_replace(ceph_disk_occupation,\"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")",
- "format": "time_series",
- "hide": false,
- "intervalFactor": 1,
- "legendFormat": "{{device}}({{osd_id}})",
- "refId": "D"
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "$ceph_hosts Disk Latency",
- "tooltip": {
- "shared": true,
- "sort": 2,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "s",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "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": "$datasource",
+ "description": "For OSD hosts, this chart shows the disk bandwidth (read bytes/sec + write bytes/sec) of the physical OSD device. Each device is shown by device name, and corresponding OSD id",
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 12,
+ "y": 12
+ },
+ "id": 8,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "/.*read/",
+ "transform": "negative-Y"
+ }
+ ],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "(irate(node_disk_bytes_written{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) write",
+ "refId": "B"
},
{
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "$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",
- "fill": 1,
- "gridPos": {
- "h": 9,
- "w": 11,
- "x": 12,
- "y": 21
- },
- "id": 8,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": true,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "minSpan": 12,
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": " ((irate(node_disk_bytes_read{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) + \n (irate(node_disk_bytes_written{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_written_bytes_total{instance=~\"($ceph_hosts).*\"}[5m]))\n ) * \n on(instance,device) group_left(osd_id) label_replace(ceph_disk_occupation,\"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")",
- "format": "time_series",
- "interval": "",
- "intervalFactor": 1,
- "legendFormat": "{{device}}({{osd_id}})",
- "refId": "A",
- "step": 10,
- "textEditor": true
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "$ceph_hosts Throughput by Disk",
- "tooltip": {
- "shared": true,
- "sort": 2,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "decbytes",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
+ "expr": "(irate(node_disk_bytes_read{instance=~\"($ceph_hosts).*\"}[5m]) or irate(node_disk_read_bytes_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{ceph_daemon}}) read",
+ "refId": "C"
}
],
- "repeat": null,
- "title": "OSD Disk Performance Statistics",
- "type": "row"
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Throughput by Disk",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "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": "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": "$datasource",
+ "description": "For OSD hosts, this chart shows the latency at the physical drive. Each drive is shown by device name, with it's corresponding OSD id",
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 0,
+ "y": 21
+ },
+ "id": 7,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "null as zero",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max by(instance,device) (((\n (irate(node_disk_write_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_writes_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n (irate(node_disk_read_time_ms{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_reads_completed{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n ) / 1000) or\n (irate(node_disk_write_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_writes_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001) or \n (irate(node_disk_read_time_seconds_total{ instance=~\"($ceph_hosts).*\"}[5m]) )\n / clamp_min(irate(node_disk_reads_completed_total{ instance=~\"($ceph_hosts).*\"}[5m]), 0.001)\n ) *\n on(instance,device) group_left(osd_id) label_replace(ceph_disk_occupation,\"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{osd_id}})",
+ "refId": "D"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk Latency",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": "0",
+ "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": "$datasource",
+ "description": "Show disk utilization % (util) of any OSD devices on the host by the physical device name and associated OSD id.",
+ "fill": 1,
+ "gridPos": {
+ "h": 9,
+ "w": 11,
+ "x": 12,
+ "y": 21
+ },
+ "id": 5,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": true,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "minSpan": 12,
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "(\n (irate(node_disk_io_time_ms[5m]) / 10 ) or\n irate(node_disk_io_time_seconds_total[5m]) * 100)\n* on(instance, device) group_left(osd_id) label_replace(ceph_disk_occupation{instance=~\"($ceph_hosts).*\"}, \"osd_id\",\"$1\",\"ceph_daemon\",\"osd.(.*)\")\n",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 1,
+ "legendFormat": "{{device}}({{osd_id}})",
+ "refId": "C",
+ "step": 10,
+ "textEditor": true
+ },
+ {
+ "expr": "(irate(node_disk_io_time_ms{instance=~\"($ceph_hosts).*\"}[5m] * 1000) or irate(node_disk_writes_completed_total{instance=~\"($ceph_hosts).*\"}[5m])) * on(instance, device) group_left(ceph_daemon) label_replace(label_replace(ceph_disk_occupation, \"device\", \"$1\", \"device\", \"/dev/(.*)\"), \"instance\", \"$1:9100\", \"exported_instance\", \"(.*)\")",
+ "format": "time_series",
+ "intervalFactor": 1,
+ "refId": "A"
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "$ceph_hosts Disk utilisation",
+ "tooltip": {
+ "shared": true,
+ "sort": 2,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "%Util",
+ "logBase": 1,
+ "max": "100",
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ],
+ "yaxis": {
+ "align": false,
+ "alignLevel": null
+ }
}
],
"refresh": "10s",
"list": [
{
"current": {
- "tags": [],
"text": "default",
"value": "default"
},
"query": "prometheus",
"refresh": 1,
"regex": "",
+ "skipUrlSync": false,
"type": "datasource"
},
{
"options": [],
"query": "label_values(node_scrape_collector_success, instance) ",
"refresh": 1,
- "regex": "([^.]*).*",
+ "regex": "([^:]*).*",
+ "skipUrlSync": false,
"sort": 3,
"tagValuesQuery": "",
"tags": [],
},
"timezone": "browser",
"title": "Host Details",
- "uid": "7IGu2Ttmz",
- "version": 11
+ "uid": "rtOg0AiWz",
+ "version": 3
}
projects / ceph.git / blobdiff