Skip to content

Commit

Permalink
Merge branch 'mjpieters-master-patch-50811' into 'master'
Browse files Browse the repository at this point in the history 
Grafana dashboard corrections

See merge request nvidia/container-toolkit/gpu-monitoring-tools!31
  • Loading branch information
@nvjmayo
nvjmayo committed Sep 23, 2020
2 parents 1e3b4db + 3439a3c commit 3a29257
Showing 1 changed file with 50 additions and 45 deletions.
95 changes: 50 additions & 45 deletions grafana/dcgm-exporter-dashboard.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -100,8 +100,8 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_GPU_TEMP",
"instant": true,
"expr": "DCGM_FI_DEV_GPU_TEMP{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"instant": false,
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand Down Expand Up @@ -130,8 +130,8 @@
"format": "celsius",
"label": null,
"logBase": 1,
"max": "100",
"min": "0",
"max": null,
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -198,7 +198,7 @@
"pluginVersion": "6.7.3",
"targets": [
{
"expr": "avg(DCGM_FI_DEV_GPU_TEMP)",
"expr": "avg(DCGM_FI_DEV_GPU_TEMP{instance=~\"${instance}\", gpu=~\"${gpu}\"})",
"interval": "",
"legendFormat": "",
"refId": "A"
Expand Down Expand Up @@ -253,7 +253,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_POWER_USAGE",
"expr": "DCGM_FI_DEV_POWER_USAGE{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand Down Expand Up @@ -283,7 +283,7 @@
"label": null,
"logBase": 1,
"max": null,
"min": "0",
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -320,24 +320,7 @@
"color": {
"mode": "thresholds"
},
"mappings": [
{
"id": 0,
"op": "=",
"text": "value to text",
"type": 1,
"value": "1"
},
{
"from": "",
"id": 1,
"operator": "",
"text": "range to text",
"to": "",
"type": 1,
"value": "2"
}
],
"mappings": [],
"max": 2400,
"min": 0,
"nullValueMode": "connected",
Expand Down Expand Up @@ -370,7 +353,7 @@
"pluginVersion": "6.7.3",
"targets": [
{
"expr": "sum(DCGM_FI_DEV_POWER_USAGE)",
"expr": "sum(DCGM_FI_DEV_POWER_USAGE{instance=~\"${instance}\", gpu=~\"${gpu}\"})",
"interval": "",
"legendFormat": "",
"refId": "A"
Expand Down Expand Up @@ -426,7 +409,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_SM_CLOCK",
"expr": "DCGM_FI_DEV_SM_CLOCK{instance=~\"${instance}\", gpu=~\"${gpu}\"} * 1000000",
"format": "time_series",
"interval": "",
"intervalFactor": 1,
Expand Down Expand Up @@ -458,8 +441,8 @@
"format": "hertz",
"label": "",
"logBase": 1,
"max": "100",
"min": "0",
"max": null,
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -519,7 +502,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_MEM_CLOCK",
"expr": "DCGM_FI_DEV_MEM_CLOCK{instance=~\"${instance}\", gpu=~\"${gpu}\"} * 1000000",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand Down Expand Up @@ -548,8 +531,8 @@
"format": "hertz",
"label": null,
"logBase": 1,
"max": "100",
"min": "0",
"max": null,
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -609,7 +592,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_GPU_UTIL",
"expr": "DCGM_FI_DEV_GPU_UTIL{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand Down Expand Up @@ -699,7 +682,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_MEM_COPY_UTIL",
"expr": "DCGM_FI_DEV_MEM_COPY_UTIL{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand Down Expand Up @@ -788,7 +771,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_FB_USED",
"expr": "DCGM_FI_DEV_FB_USED{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand All @@ -814,11 +797,11 @@
},
"yaxes": [
{
"format": "decbytes",
"format": "decmbytes",
"label": null,
"logBase": 1,
"max": "100",
"min": "0",
"max": null,
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -878,7 +861,7 @@
"steppedLine": false,
"targets": [
{
"expr": "DCGM_FI_DEV_FB_USED",
"expr": "DCGM_FI_DEV_FB_FREE{instance=~\"${instance}\", gpu=~\"${gpu}\"}",
"interval": "",
"legendFormat": "GPU {{gpu}}",
"refId": "A"
Expand All @@ -904,11 +887,11 @@
},
"yaxes": [
{
"format": "decbytes",
"format": "decmbytes",
"label": null,
"logBase": 1,
"max": "100",
"min": "0",
"max": null,
"min": null,
"show": true
},
{
Expand Down Expand Up @@ -936,12 +919,34 @@
"allValue": null,
"current": {},
"datasource": "${DS_PROMETHEUS}",
"definition": "label_values(gpu)",
"definition": "label_values(DCGM_FI_DEV_GPU_TEMP, instance)",
"hide": 0,
"includeAll": false,
"label": null,
"multi": true,
"name": "instance",
"options": [],
"query": "label_values(DCGM_FI_DEV_GPU_TEMP, instance)",
"refresh": 1,
"regex": "",
"skipUrlSync": false,
"sort": 0,
"tagValuesQuery": "",
"tags": [],
"tagsQuery": "",
"type": "query",
"useTags": false
},
{
"allValue": null,
"current": {},
"datasource": "${DS_PROMETHEUS}",
"definition": "label_values(gpu)",
"hide": 0,
"includeAll": true,
"index": -1,
"label": null,
"multi": false,
"multi": true,
"name": "gpu",
"options": [],
"query": "label_values(gpu)",
Expand Down Expand Up @@ -982,4 +987,4 @@
"list": []
},
"version": 1
}
}

0 comments on commit 3a29257

Please sign in to comment.