diff --git a/grafana/dashboards/health.json b/grafana/dashboards/health.json index 64b30031c..374b2f3ca 100644 --- a/grafana/dashboards/health.json +++ b/grafana/dashboards/health.json @@ -16,16 +16,305 @@ "gnetId": null, "graphTooltip": 1, "id": null, - "iteration": 1524697499292, + "iteration": 1529950676425, "links": [], "panels": [ + { + "content": "
\n Data-Plane Telemetry\n
", + "gridPos": { + "h": 2.2, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 400, + "links": [], + "mode": "html", + "title": "", + "transparent": true, + "type": "text" + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "fill": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 2.2 + }, + "id": 397, + "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": "process_virtual_memory_bytes{conduit_io_control_plane_ns=\"$namespace\", job=\"conduit-proxy\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}/{{pod}}/virtual", + "refId": "A" + }, + { + "expr": "process_resident_memory_bytes{conduit_io_control_plane_ns=\"$namespace\", job=\"conduit-proxy\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}/{{pod}}/resident", + "refId": "B" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "MEMORY USAGE", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "decbytes", + "label": null, + "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": "prometheus", + "fill": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 2.2 + }, + "id": 399, + "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(process_cpu_seconds_total{conduit_io_control_plane_ns=\"$namespace\", job=\"conduit-proxy\"}[20s])) by (namespace, pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}/{{pod}}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "CPU USAGE", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": "0", + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "prometheus", + "fill": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 2.2 + }, + "id": 398, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_open_fds{conduit_io_control_plane_ns=\"$namespace\", job=\"conduit-proxy\"}", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}/{{pod}}", + "refId": "A" + } + ], + "thresholds": [], + "timeFrom": null, + "timeShift": null, + "title": "OPEN FILE DESCRIPTORS", + "tooltip": { + "shared": false, + "sort": 2, + "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": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } + }, + { + "content": "
\n Control-Plane Traffic\n
", + "gridPos": { + "h": 2.2, + "w": 24, + "x": 0, + "y": 9.2 + }, + "id": 401, + "links": [], + "mode": "html", + "title": "", + "transparent": true, + "type": "text" + }, { "collapsed": false, "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 0 + "y": 11.4 }, "id": 19, "panels": [], @@ -39,7 +328,7 @@ "h": 2.2, "w": 24, "x": 0, - "y": 1 + "y": 12.4 }, "id": 21, "links": [], @@ -59,7 +348,7 @@ "h": 7, "w": 8, "x": 0, - "y": 3.2 + "y": 14.6 }, "id": 23, "legend": { @@ -143,7 +432,7 @@ "h": 7, "w": 8, "x": 8, - "y": 3.2 + "y": 14.6 }, "id": 24, "legend": { @@ -227,7 +516,7 @@ "h": 7, "w": 8, "x": 16, - "y": 3.2 + "y": 14.6 }, "id": 25, "legend": { @@ -320,7 +609,7 @@ "h": 1, "w": 24, "x": 0, - "y": 10.2 + "y": 21.6 }, "id": 339, "panels": [], @@ -333,7 +622,7 @@ "h": 2.2, "w": 24, "x": 0, - "y": 11.2 + "y": 22.6 }, "id": 340, "links": [], @@ -348,7 +637,7 @@ "h": 1, "w": 24, "x": 0, - "y": 13.4 + "y": 24.8 }, "id": 179, "panels": [ @@ -612,7 +901,7 @@ "h": 1, "w": 24, "x": 0, - "y": 14.4 + "y": 25.8 }, "id": 90, "panels": [], @@ -625,7 +914,7 @@ "h": 2.2, "w": 24, "x": 0, - "y": 15.4 + "y": 26.8 }, "id": 27, "links": [], @@ -645,7 +934,7 @@ "h": 7, "w": 8, "x": 0, - "y": 17.6 + "y": 29 }, "id": 2, "legend": { @@ -729,7 +1018,7 @@ "h": 7, "w": 8, "x": 8, - "y": 17.6 + "y": 29 }, "id": 5, "legend": { @@ -758,14 +1047,14 @@ "expr": "process_resident_memory_bytes{job=\"conduit-controller\"}", "format": "time_series", "intervalFactor": 1, - "legendFormat": "resident/{{component}}", + "legendFormat": "{{component}}/resident", "refId": "A" }, { "expr": "process_virtual_memory_bytes{job=\"conduit-controller\"}", "format": "time_series", "intervalFactor": 1, - "legendFormat": "virtual/{{component}}", + "legendFormat": "{{component}}/virtual", "refId": "B" } ], @@ -820,7 +1109,7 @@ "h": 7, "w": 8, "x": 16, - "y": 17.6 + "y": 29 }, "id": 7, "legend": { @@ -904,7 +1193,7 @@ "h": 7, "w": 8, "x": 0, - "y": 24.6 + "y": 36 }, "id": 9, "legend": { @@ -988,7 +1277,7 @@ "h": 7, "w": 8, "x": 8, - "y": 24.6 + "y": 36 }, "id": 12, "legend": { @@ -1081,7 +1370,7 @@ "h": 1, "w": 24, "x": 0, - "y": 31.6 + "y": 43 }, "id": 29, "panels": [], @@ -1095,7 +1384,7 @@ "h": 2.2, "w": 24, "x": 0, - "y": 32.6 + "y": 44 }, "id": 30, "links": [], @@ -1115,7 +1404,7 @@ "h": 7, "w": 8, "x": 0, - "y": 34.8 + "y": 46.2 }, "id": 6, "legend": { @@ -1220,7 +1509,7 @@ "h": 7, "w": 8, "x": 8, - "y": 34.8 + "y": 46.2 }, "id": 8, "legend": { @@ -1318,7 +1607,7 @@ "h": 7, "w": 8, "x": 16, - "y": 34.8 + "y": 46.2 }, "id": 14, "legend": {