diff --git a/dashboards/MongoDB/MongoDB_ReplSet_Summary.json b/dashboards/MongoDB/MongoDB_ReplSet_Summary.json index 066e4ccc3d..678f58083d 100644 --- a/dashboards/MongoDB/MongoDB_ReplSet_Summary.json +++ b/dashboards/MongoDB/MongoDB_ReplSet_Summary.json @@ -3,7 +3,10 @@ "list": [ { "builtIn": 1, - "datasource": "-- Grafana --", + "datasource": { + "type": "datasource", + "uid": "grafana" + }, "enable": true, "hide": false, "iconColor": "#e0752d", @@ -28,7 +31,10 @@ }, { "builtIn": 1, - "datasource": "-- Grafana --", + "datasource": { + "type": "datasource", + "uid": "grafana" + }, "enable": true, "hide": true, "iconColor": "#6ed0e0", @@ -49,8 +55,6 @@ "editable": false, "fiscalYearStartMonth": 0, "graphTooltip": 1, - "id": null, - "iteration": 1656496368829, "links": [ { "icon": "doc", @@ -137,6 +141,7 @@ "panels": [ { "collapsed": false, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, @@ -145,10 +150,17 @@ }, "id": 1022, "panels": [], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], "title": "Overview", "type": "row" }, { + "datasource": "Metrics", "description": "", "gridPos": { "h": 3, @@ -159,13 +171,19 @@ "id": 1228, "links": [], "options": { + "code": { + "language": "plaintext", + "showLineNumbers": false, + "showMiniMap": false + }, "content": "
$cluster
", "mode": "html" }, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "targets": [ { "calculatedInterval": "10m", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "mysql_global_variables_innodb_buffer_pool_size{service_name=~\"$service_name\"} ", @@ -182,6 +200,7 @@ "type": "text" }, { + "datasource": "Metrics", "description": "This shows how many members are configured in the replica set.", "fieldConfig": { "defaults": { @@ -235,9 +254,10 @@ }, "textMode": "auto" }, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "targets": [ { + "datasource": "Metrics", "expr": "count by (set) (group by (service_name, set) (mongodb_mongod_replset_number_of_members{cluster=~\"$cluster\",set=~\"$replset\"} or mongodb_mongod_replset_my_state{cluster=~\"$cluster\",set=~\"$replset\"}))", "interval": "5m", "intervalFactor": 1, @@ -251,6 +271,7 @@ "type": "stat" }, { + "datasource": "Metrics", "description": "This shows the time since the last election.", "fieldConfig": { "defaults": { @@ -304,9 +325,10 @@ }, "textMode": "auto" }, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "targets": [ { + "datasource": "Metrics", "expr": "time() - max(mongodb_mongod_replset_member_election_date{service_name=~\"$service_name\"})", "interval": "5m", "intervalFactor": 1, @@ -320,6 +342,7 @@ "type": "stat" }, { + "datasource": "Metrics", "description": "This panel shows how far behind in replication this member is if it is a secondary. This number may be high it the instance is running as a delayed secondary member.", "fieldConfig": { "defaults": { @@ -373,7 +396,7 @@ }, "textMode": "auto" }, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "targets": [ { "datasource": "Metrics", @@ -393,6 +416,7 @@ }, { "columns": [], + "datasource": "Metrics", "description": "", "fontSize": "100%", "gridPos": { @@ -502,32 +526,21 @@ "type": "table-old" }, { - "id": null, - "gridPos": { - "h": 8, - "w": 17, - "x": 0, - "y": 4 - }, - "type": "state-timeline", - "title": "ReplSet States", - "transformations": [], - "pluginVersion": "6.5.1", + "datasource": "Metrics", "description": "ReplSet statuses during the select time range.", "fieldConfig": { "defaults": { + "color": { + "fixedColor": "blue", + "mode": "palette-classic" + }, "custom": { - "lineWidth": 0, "fillOpacity": 100, + "lineWidth": 0, "spanNulls": false }, - "color": { - "mode": "palette-classic", - "fixedColor": "blue" - }, "mappings": [ { - "type": "value", "options": { "0": { "index": 11, @@ -577,18 +590,19 @@ "index": 2, "text": "Exporter is not connected" } - } + }, + "type": "value" }, { - "type": "range", "options": { "from": 0, - "to": 1, "result": { "index": 0, "text": "N/A" - } - } + }, + "to": 1 + }, + "type": "range" } ], "thresholds": { @@ -608,23 +622,32 @@ }, "overrides": [] }, + "gridPos": { + "h": 8, + "w": 17, + "x": 0, + "y": 4 + }, + "id": 1559, "options": { - "mergeValues": true, - "showValue": "auto", "alignValue": "right", - "rowHeight": 0.9, "legend": { - "showLegend": true, "displayMode": "list", - "placement": "bottom" + "placement": "bottom", + "showLegend": true }, + "mergeValues": true, + "rowHeight": 0.9, + "showValue": "auto", "tooltip": { "mode": "single", "sort": "none" } }, + "pluginVersion": "6.5.1", "targets": [ { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "max by (service_name) (mongodb_mongod_replset_my_state{set=~\"$replset\",service_name=~\"$service_name\"})", @@ -636,10 +659,14 @@ "range": true, "refId": "A" } - ] + ], + "title": "ReplSet States", + "transformations": [], + "type": "state-timeline" }, { - "collapsed": false, + "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, @@ -647,841 +674,900 @@ "y": 12 }, "id": 1024, - "panels": [], - "title": "Replication Lag", - "type": "row" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "MongoDB replication lag occurs when the secondary node cannot replicate data fast enough to keep up with the rate that data is being written to the primary node. It could be caused by something as simple as network latency, packet loss within your network, or a routing issue.", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { - "links": [ + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", + "decimals": 2, + "description": "MongoDB replication lag occurs when the secondary node cannot replicate data fast enough to keep up with the rate that data is being written to the primary node. It could be caused by something as simple as network latency, packet loss within your network, or a routing issue.", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__series.name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 24, + "x": 0, + "y": 34 + }, + "hiddenSeries": false, + "id": 14, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sort": "avg", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null as zero", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ { - "targetBlank": true, - "title": "MongoDB Instance Summary - ${__series.name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" + "alias": "Avg", + "color": "#C4162A", + "fill": 0, + "legend": false, + "stack": false } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 24, - "x": 0, - "y": 13 - }, - "hiddenSeries": false, - "id": 14, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "max": true, - "min": true, - "rightSide": true, - "show": true, - "sort": "avg", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null as zero", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - { - "alias": "Avg", - "color": "#C4162A", - "fill": 0, - "legend": false, - "stack": false + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "avg by (service_name) ((max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[$interval]) > 0) by (service_name,set)) or (max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[5m]) > 0) by (service_name,set)))", + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "{{service_name}}", + "refId": "A", + "step": 300 + }, + { + "datasource": "Metrics", + "expr": "avg by (service_name) ((max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[$interval]) > 0) by (service_name,set)) or (max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[5m]) > 0) by (service_name,set)))", + "hide": true, + "interval": "$interval", + "legendFormat": "Avg", + "refId": "B" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Replication Lag", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "cumulative" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "s", + "short" + ], + "yaxes": [ + { + "decimals": 2, + "format": "s", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } } ], - "spaceLength": 10, - "stack": false, - "steppedLine": false, "targets": [ { - "expr": "avg by (service_name) ((max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[$interval]) > 0) by (service_name,set)) or (max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[5m]) > 0) by (service_name,set)))", - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "{{service_name}}", - "refId": "A", - "step": 300 - }, - { - "expr": "avg by (service_name) ((max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[$interval]) > 0) by (service_name,set)) or (max(max_over_time(mongodb_mongod_replset_member_replication_lag{set=\"$replset\",state=\"SECONDARY\",self=~\"|1\",service_name=~\"$secondary\"}[5m]) > 0) by (service_name,set)))", - "hide": true, - "interval": "$interval", - "legendFormat": "Avg", - "refId": "B" + "datasource": "Metrics", + "refId": "A" } ], - "thresholds": [], - "timeRegions": [], "title": "Replication Lag", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "cumulative" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "y-axis": true, - "y_formats": [ - "s", - "short" - ], - "yaxes": [ - { - "decimals": 2, - "format": "s", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": false - } - ], - "yaxis": { - "align": false - } + "type": "row" }, { - "collapsed": false, + "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 21 + "y": 13 }, "id": 1019, - "panels": [], - "title": "Operations", - "type": "row" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "Operations are classified by legacy wire protocol type (insert, update, and delete only).", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { - "links": [ - { - "targetBlank": true, - "title": "MongoDB Instance Summary - ${__field.labels.service_name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" - } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 24, - "x": 0, - "y": 22 - }, - "hiddenSeries": false, - "id": 1020, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "max": true, - "min": true, - "show": true, - "sort": "avg", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [ - { - "targetBlank": true, - "title": "MongoDB Instance Summary - $service_name", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=$service_name$__url_time_range" - } - ], - "nullPointMode": "null", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": "service_name", - "repeatDirection": "h", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "avg by (service_name,type) (rate(mongodb_op_counters_repl_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_op_counters_repl_total{service_name=~\"$service_name\"}[5m]))", - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "repl - {{type}}", - "refId": "A", - "step": 300 - }, - { - "expr": "avg by (service_name,type) (rate(mongodb_mongod_op_counters_repl_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_op_counters_repl_total{service_name=~\"$service_name\"}[5m]))", - "interval": "$interval", - "legendFormat": "repl - {{type}}", - "refId": "B" - }, - { - "expr": "avg by (service_name,type) (rate(mongodb_op_counters_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_op_counters_total{service_name=~\"$service_name\"}[5m]))", - "interval": "$interval", - "legendFormat": "{{type}}", - "refId": "C" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Operations - $service_name", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "cumulative" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "y-axis": true, - "y_formats": [ - "short", - "short" - ], - "yaxes": [ + "panels": [ { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", "decimals": 2, - "format": "ops", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": false - } - ], - "yaxis": { - "align": false - } - }, - { - "collapsed": false, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 30 - }, - "id": 1170, - "panels": [], - "title": "Max Member Ping Time", - "type": "row" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "This metric can show a correlation with the replication lag value.", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { + "description": "Operations are classified by legacy wire protocol type (insert, update, and delete only).", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__field.labels.service_name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 0, + "y": 35 + }, + "hiddenSeries": false, + "id": 1020, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "show": true, + "sort": "avg", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, "links": [ { "targetBlank": true, - "title": "MongoDB Instance Summary - ${__field.labels.service_name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" + "title": "MongoDB Instance Summary - $service_name", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=$service_name$__url_time_range" } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 24, - "x": 0, - "y": 31 - }, - "hiddenSeries": false, - "id": 13, - "legend": { - "alignAsTable": true, - "avg": false, - "current": false, - "max": true, - "min": false, - "show": true, - "sort": "max", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": "service_name", - "repeatDirection": "h", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "avg by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{service_name=~\"$service_name\"})\n* on (name) group_right(state) avg by (service_name,state,name) ((max_over_time(mongodb_mongod_replset_my_name[$interval]) or max_over_time(mongodb_mongod_replset_my_name[5m])))", - "format": "time_series", - "hide": true, - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "{{service_name}} - {{name}} - {{state}}", - "refId": "A", - "step": 300 - }, - { - "expr": "max by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{set=~\"$replset\",service_name=~\"$service_name\"}) or mongodb_rs_members_pingMs{service_name=~\"$service_name\", member_state!=\"\"}", - "hide": true, - "interval": "$interval", - "legendFormat": "{{service_name}} - {{name}} - {{state}}", - "refId": "B" - }, - { - "expr": "max by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{set=~\"$replset\",service_name=~\"$service_name\"} or label_replace(label_replace(mongodb_rs_members_pingMs{service_name=~\"$service_name\", member_state!=\"\"},\"state\", \"$1\", \"member_state\", \"(.*)\"),\"name\", \"$1\", \"member_idx\", \"(.*)\"))", - "interval": "$interval", - "legendFormat": "{{service_name}} - {{name}}{{member_idx}} - {{state}}{{member_state}}", - "refId": "D" + ], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": "service_name", + "repeatDirection": "h", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "avg by (service_name,type) (rate(mongodb_op_counters_repl_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_op_counters_repl_total{service_name=~\"$service_name\"}[5m]))", + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "repl - {{type}}", + "refId": "A", + "step": 300 + }, + { + "datasource": "Metrics", + "expr": "avg by (service_name,type) (rate(mongodb_mongod_op_counters_repl_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_op_counters_repl_total{service_name=~\"$service_name\"}[5m]))", + "interval": "$interval", + "legendFormat": "repl - {{type}}", + "refId": "B" + }, + { + "datasource": "Metrics", + "expr": "avg by (service_name,type) (rate(mongodb_op_counters_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_op_counters_total{service_name=~\"$service_name\"}[5m]))", + "interval": "$interval", + "legendFormat": "{{type}}", + "refId": "C" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Operations - $service_name", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "cumulative" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "short", + "short" + ], + "yaxes": [ + { + "decimals": 2, + "format": "ops", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } } ], - "thresholds": [], - "timeRegions": [], - "title": "Max Member Ping Time - $service_name", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "cumulative" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "y-axis": true, - "y_formats": [ - "ms", - "short" - ], - "yaxes": [ - { - "decimals": 2, - "format": "ms", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, + "targets": [ { - "format": "short", - "logBase": 1, - "show": false + "datasource": "Metrics", + "refId": "A" } ], - "yaxis": { - "align": false - } + "title": "Operations", + "type": "row" }, { - "collapsed": false, + "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 39 - }, - "id": 1017, - "panels": [], - "title": "Max Heartbeat Time / Elections", - "type": "row" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "Time span between now and last heartbeat from replicaset members.", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { - "links": [ - { - "targetBlank": true, - "title": "MongoDB Instance Summary - ${__series.name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" - } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 12, - "x": 0, - "y": 40 - }, - "hiddenSeries": false, - "id": 75, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "max": true, - "min": true, - "rightSide": false, - "show": true, - "sort": "avg", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "time() - avg by (service_name) (max(mongodb_mongod_replset_member_last_heartbeat{service_name=~\"$service_name\"}) by (name)) * on (name) group_right avg by (service_name) (mongodb_mongod_replset_my_name{service_name=~\"$service_name\"})", - "hide": false, - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "{{service_name}}", - "metric": "", - "refId": "J", - "step": 300 - }, - { - "expr": "avg by (service_name) (max(mongodb_rs_heartbeatIntervalMillis{service_name=~\"$service_name\"}) by (name) / 1000) * on (name) group_right avg by (service_name) (mongodb_mongod_replset_my_name{service_name=~\"$service_name\"})", - "hide": true, - "interval": "$interval", - "legendFormat": "Interval - {service_name}}", - "refId": "B" - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Max Heartbeat Time", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "individual" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] + "y": 14 }, - "y-axis": true, - "y_formats": [ - "s", - "short" - ], - "yaxes": [ + "id": 1170, + "panels": [ { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", "decimals": 2, - "format": "s", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": false - } - ], - "yaxis": { - "align": false - } - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "Count of elections. Usually zero; 1 count by each healthy node will appear in each election. Happens when the primary role changes due to either normal maintenance or trouble events.", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { - "links": [ + "description": "This metric can show a correlation with the replication lag value.", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__field.labels.service_name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 0, + "y": 36 + }, + "hiddenSeries": false, + "id": 13, + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": true, + "min": false, + "show": true, + "sort": "max", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": "service_name", + "repeatDirection": "h", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "avg by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{service_name=~\"$service_name\"})\n* on (name) group_right(state) avg by (service_name,state,name) ((max_over_time(mongodb_mongod_replset_my_name[$interval]) or max_over_time(mongodb_mongod_replset_my_name[5m])))", + "format": "time_series", + "hide": true, + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "{{service_name}} - {{name}} - {{state}}", + "refId": "A", + "step": 300 + }, + { + "datasource": "Metrics", + "expr": "max by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{set=~\"$replset\",service_name=~\"$service_name\"}) or mongodb_rs_members_pingMs{service_name=~\"$service_name\", member_state!=\"\"}", + "hide": true, + "interval": "$interval", + "legendFormat": "{{service_name}} - {{name}} - {{state}}", + "refId": "B" + }, + { + "datasource": "Metrics", + "expr": "max by (service_name,name,state) (mongodb_mongod_replset_member_ping_ms{set=~\"$replset\",service_name=~\"$service_name\"} or label_replace(label_replace(mongodb_rs_members_pingMs{service_name=~\"$service_name\", member_state!=\"\"},\"state\", \"$1\", \"member_state\", \"(.*)\"),\"name\", \"$1\", \"member_idx\", \"(.*)\"))", + "interval": "$interval", + "legendFormat": "{{service_name}} - {{name}}{{member_idx}} - {{state}}{{member_state}}", + "refId": "D" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Max Member Ping Time - $service_name", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "cumulative" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "ms", + "short" + ], + "yaxes": [ + { + "decimals": 2, + "format": "ms", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], + "title": "Max Member Ping Time", + "type": "row" + }, + { + "collapsed": true, + "datasource": "Metrics", + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 15 + }, + "id": 1017, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", + "decimals": 2, + "description": "Time span between now and last heartbeat from replicaset members.", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__series.name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 0, + "y": 37 + }, + "hiddenSeries": false, + "id": 75, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "rightSide": false, + "show": true, + "sort": "avg", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "time() - avg by (service_name) (max(mongodb_mongod_replset_member_last_heartbeat{service_name=~\"$service_name\"}) by (name)) * on (name) group_right avg by (service_name) (mongodb_mongod_replset_my_name{service_name=~\"$service_name\"})", + "hide": false, + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "{{service_name}}", + "metric": "", + "refId": "J", + "step": 300 + }, + { + "datasource": "Metrics", + "expr": "avg by (service_name) (max(mongodb_rs_heartbeatIntervalMillis{service_name=~\"$service_name\"}) by (name) / 1000) * on (name) group_right avg by (service_name) (mongodb_mongod_replset_my_name{service_name=~\"$service_name\"})", + "hide": true, + "interval": "$interval", + "legendFormat": "Interval - {service_name}}", + "refId": "B" + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Max Heartbeat Time", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "individual" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "s", + "short" + ], + "yaxes": [ + { + "decimals": 2, + "format": "s", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + }, + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", + "decimals": 2, + "description": "Count of elections. Usually zero; 1 count by each healthy node will appear in each election. Happens when the primary role changes due to either normal maintenance or trouble events.", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__series.name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 12, + "x": 12, + "y": 37 + }, + "hiddenSeries": false, + "id": 12, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "show": true, + "sort": "avg", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "max by (service_name) (changes(mongodb_mongod_replset_member_election_date{service_name=~\"$service_name\"}[$interval]))", + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "{{service_name}}", + "refId": "A", + "step": 300 + } + ], + "thresholds": [], + "timeRegions": [], + "title": "Elections", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "cumulative" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "short", + "short" + ], + "yaxes": [ + { + "decimals": 2, + "format": "short", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "logBase": 1, + "show": false + } + ], + "yaxis": { + "align": false + } + } + ], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], + "title": "Max Heartbeat Time / Elections", + "type": "row" + }, + { + "collapsed": true, + "datasource": "Metrics", + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 16 + }, + "id": 1547, + "panels": [ + { + "aliasColors": {}, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "Metrics", + "decimals": 2, + "description": "Timespan 'window' between newest and the oldest op in the Oplog collection.", + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "links": [ + { + "targetBlank": true, + "title": "MongoDB Instance Summary - ${__field.labels.service_name}", + "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" + } + ] + }, + "overrides": [] + }, + "fill": 2, + "fillGradient": 0, + "grid": { + "leftLogBase": 1, + "leftMin": 0, + "rightLogBase": 1 + }, + "gridPos": { + "h": 8, + "w": 8, + "x": 0, + "y": 38 + }, + "hiddenSeries": false, + "id": 27, + "legend": { + "alignAsTable": true, + "avg": true, + "current": false, + "max": true, + "min": true, + "show": true, + "sort": "avg", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [], + "nullPointMode": "null", + "options": { + "alertThreshold": true + }, + "paceLength": 10, + "percentage": false, + "pluginVersion": "9.2.20", + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": "service_name", + "repeatDirection": "h", + "seriesOverrides": [ + { + "alias": "Oplog Range", + "yaxis": 2 + } + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "datasource": "Metrics", + "expr": "time()-avg by (service_name) (mongodb_mongod_replset_oplog_tail_timestamp{service_name=~\"$service_name\"})", + "hide": false, + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "Now to End", + "metric": "", + "refId": "J", + "step": 300 + }, { - "targetBlank": true, - "title": "MongoDB Instance Summary - ${__series.name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__series.name}$&$__url_time_range" + "datasource": "Metrics", + "expr": "avg by (service_name) (mongodb_mongod_replset_oplog_head_timestamp{service_name=~\"$service_name\"}-mongodb_mongod_replset_oplog_tail_timestamp{service_name=~\"$service_name\"})", + "hide": false, + "interval": "$interval", + "intervalFactor": 1, + "legendFormat": "Oplog Range", + "metric": "", + "refId": "A", + "step": 300 } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 12, - "x": 12, - "y": 40 - }, - "hiddenSeries": false, - "id": 12, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "max": true, - "min": true, - "show": true, - "sort": "avg", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "max by (service_name) (changes(mongodb_mongod_replset_member_election_date{service_name=~\"$service_name\"}[$interval]))", - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "{{service_name}}", - "refId": "A", - "step": 300 - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Elections", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "cumulative" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "y-axis": true, - "y_formats": [ - "short", - "short" - ], - "yaxes": [ - { - "decimals": 2, - "format": "short", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "logBase": 1, - "show": false - } - ], - "yaxis": { - "align": false - } - }, - { - "collapsed": false, - "gridPos": { - "h": 1, - "w": 24, - "x": 0, - "y": 48 - }, - "id": 1547, - "panels": [], - "title": "Oplog Recovery Window", - "type": "row" - }, - { - "aliasColors": {}, - "bars": false, - "dashLength": 10, - "dashes": false, - "decimals": 2, - "description": "Timespan 'window' between newest and the oldest op in the Oplog collection.", - "editable": true, - "error": false, - "fieldConfig": { - "defaults": { - "links": [ + ], + "thresholds": [], + "timeRegions": [], + "title": "Oplog Recovery Window - $service_name", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 5, + "value_type": "cumulative" + }, + "type": "graph", + "x-axis": true, + "xaxis": { + "mode": "time", + "show": true, + "values": [] + }, + "y-axis": true, + "y_formats": [ + "s", + "short" + ], + "yaxes": [ { - "targetBlank": true, - "title": "MongoDB Instance Summary - ${__field.labels.service_name}", - "url": "/graph/d/mongodb-instance-summary/mongodb-instance-summary?var-service_name=${__field.labels.service_name}&$__url_time_range" + "decimals": 2, + "format": "s", + "label": "", + "logBase": 1, + "min": 0, + "show": true + }, + { + "format": "s", + "logBase": 1, + "show": true } - ] - }, - "overrides": [] - }, - "fill": 2, - "fillGradient": 0, - "grid": { - "leftLogBase": 1, - "leftMin": 0, - "rightLogBase": 1 - }, - "gridPos": { - "h": 8, - "w": 24, - "x": 0, - "y": 49 - }, - "hiddenSeries": false, - "id": 27, - "legend": { - "alignAsTable": true, - "avg": true, - "current": false, - "max": true, - "min": true, - "show": true, - "sort": "avg", - "sortDesc": true, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 2, - "links": [], - "nullPointMode": "null", - "options": { - "alertThreshold": true - }, - "paceLength": 10, - "percentage": false, - "pluginVersion": "8.3.5", - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": "service_name", - "repeatDirection": "h", - "seriesOverrides": [ - { - "alias": "Oplog Range", - "yaxis": 2 + ], + "yaxis": { + "align": false + } } ], - "spaceLength": 10, - "stack": false, - "steppedLine": false, "targets": [ { - "expr": "time()-avg by (service_name) (mongodb_mongod_replset_oplog_tail_timestamp{service_name=~\"$service_name\"})", - "hide": false, - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "Now to End", - "metric": "", - "refId": "J", - "step": 300 - }, - { - "expr": "avg by (service_name) (mongodb_mongod_replset_oplog_head_timestamp{service_name=~\"$service_name\"}-mongodb_mongod_replset_oplog_tail_timestamp{service_name=~\"$service_name\"})", - "hide": false, - "interval": "$interval", - "intervalFactor": 1, - "legendFormat": "Oplog Range", - "metric": "", - "refId": "A", - "step": 300 - } - ], - "thresholds": [], - "timeRegions": [], - "title": "Oplog Recovery Window - $service_name", - "tooltip": { - "msResolution": false, - "shared": true, - "sort": 5, - "value_type": "cumulative" - }, - "type": "graph", - "x-axis": true, - "xaxis": { - "mode": "time", - "show": true, - "values": [] - }, - "y-axis": true, - "y_formats": [ - "s", - "short" - ], - "yaxes": [ - { - "decimals": 2, - "format": "s", - "label": "", - "logBase": 1, - "min": 0, - "show": true - }, - { - "format": "s", - "logBase": 1, - "show": true + "datasource": "Metrics", + "refId": "A" } ], - "yaxis": { - "align": false - } + "title": "Oplog Recovery Window", + "type": "row" }, { "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 57 + "y": 17 }, "id": 1070, "panels": [ @@ -1490,6 +1576,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Repl buffer ops applied per sec.", "editable": true, @@ -1517,7 +1604,7 @@ "h": 8, "w": 12, "x": 0, - "y": 60 + "y": 100 }, "hiddenSeries": false, "id": 85, @@ -1553,6 +1640,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_mongod_metrics_repl_buffer_count{service_name=~\"$service_name\"})", "interval": "$interval", "intervalFactor": 1, @@ -1606,6 +1694,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Time spent per second waiting for or fetching oplog docs in replication.", "editable": true, @@ -1633,7 +1722,7 @@ "h": 8, "w": 12, "x": 12, - "y": 60 + "y": 100 }, "hiddenSeries": false, "id": 79, @@ -1669,6 +1758,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_network_getmores_total_milliseconds{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_network_getmores_total_milliseconds{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -1722,6 +1812,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Times spent per second A) pre-loading oplog ops into parallel-executable batches B) Times spent pre-loading index values and C) repl batch apply phase.", "editable": true, @@ -1749,7 +1840,7 @@ "h": 8, "w": 24, "x": 0, - "y": 68 + "y": 108 }, "hiddenSeries": false, "id": 84, @@ -1786,6 +1877,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_preload_docs_total_milliseconds{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_preload_docs_total_milliseconds{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -1794,6 +1886,7 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_preload_indexes_total_milliseconds{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_preload_indexes_total_milliseconds{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -1803,6 +1896,7 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_apply_batches_total_milliseconds{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_apply_batches_total_milliseconds{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -1857,6 +1951,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Current hard-coded max and actual size of repl batch buffer.", "editable": true, @@ -1884,7 +1979,7 @@ "h": 8, "w": 24, "x": 0, - "y": 76 + "y": 116 }, "hiddenSeries": false, "id": 80, @@ -1927,6 +2022,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_mongod_metrics_repl_buffer_size_bytes{service_name=~\"$service_name\"})", "interval": "$interval", "intervalFactor": 1, @@ -1935,6 +2031,7 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_mongod_metrics_repl_buffer_max_size_bytes{service_name=~\"$service_name\"})", "interval": "$interval", "intervalFactor": 1, @@ -1988,6 +2085,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Count of A) getmores executed B) index values pre-loaded C) oplog docs applied per second against oplog collection in replication.", "editable": true, @@ -2015,7 +2113,7 @@ "h": 8, "w": 24, "x": 0, - "y": 84 + "y": 124 }, "hiddenSeries": false, "id": 81, @@ -2052,6 +2150,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_preload_docs_num_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_preload_docs_num_total{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -2061,6 +2160,7 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_preload_indexes_num_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_preload_indexes_num_total{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 1, @@ -2070,6 +2170,7 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (rate(mongodb_mongod_metrics_repl_apply_ops_total{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_mongod_metrics_repl_apply_ops_total{service_name=~\"$service_name\"}[5m]))", "interval": "$interval", "intervalFactor": 2, @@ -2119,40 +2220,43 @@ } } ], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], "title": "Oplog Details", "type": "row" }, { "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 58 + "y": 18 }, "id": 1306, "panels": [ { "collapsed": false, - "gridPos": { - "h": 10, - "w": 24, - "x": 0, - "y": 61 - }, - "id": 1352, - "pluginVersion": "9.2.20", + "datasource": "Metrics", "default_title_for_rows": "Service Name", - "title": "Services Details", - "type": "table", "fieldConfig": { "defaults": { + "color": { + "fixedColor": "text", + "mode": "thresholds" + }, "custom": { "align": "center", "displayMode": "auto", - "inspect": false, - "filterable": false + "filterable": false, + "inspect": false }, + "decimals": 2, "mappings": [ { "options": { @@ -2169,25 +2273,19 @@ "type": "value" } ], + "noValue": "-", "thresholds": { "mode": "absolute", "steps": [ { - "color": "green", - "value": null + "color": "green" }, { "color": "red", "value": 80 } ] - }, - "color": { - "fixedColor": "text", - "mode": "thresholds" - }, - "noValue": "-", - "decimals": 2 + } }, "overrides": [ { @@ -2263,55 +2361,27 @@ } ] }, + "gridPos": { + "h": 10, + "w": 24, + "x": 0, + "y": 101 + }, + "id": 1352, "options": { - "showHeader": true, "footer": { "show": false - } - }, - "transformations": [ - { - "id": "merge", - "options": {} }, - { - "id": "organize", - "options": { - "excludeByName": { - "Time": true, - "agent_id": true, - "agent_type": true, - "environment": true, - "instance": true, - "job": true, - "machine_id": true, - "node_id": true, - "node_name": true, - "node_type": true, - "service_name": false - }, - "indexByName": { - "Time": 0, - "Value #A": 6, - "Value #B": 4, - "Value #C": 3, - "Value #D": 5, - "Value #E": 2, - "service_name": 1 - }, - "renameByName": { - "Value #A": "QPS", - "Value #B": "Average Latency", - "Value #C": "Uptime", - "Value #D": "Open Connections", - "Value #E": "Cursors", - "service_name": "Service Name" - } - } - } - ], + "showHeader": true + }, + "pluginVersion": "9.2.20", + "sorting_props": { + "col_index": -1, + "direction": "desc" + }, "targets": [ { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (service_name) (mongodb_instance_uptime_seconds{service_name=~\"$service_name\"})", @@ -2323,6 +2393,7 @@ "refId": "C" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "sum by (service_name) (rate(mongodb_mongod_op_counters_total{service_name=~\"$service_name\",type!=\"command\"}[$interval]) or irate(mongodb_mongod_op_counters_total{service_name=~\"$service_name\",type!=\"command\"}[5m]) or rate(mongodb_op_counters_total{service_name=~\"$service_name\",type!=\"command\"}[$interval]) or irate(mongodb_op_counters_total{service_name=~\"$service_name\",type!=\"command\"}[5m]))", @@ -2335,6 +2406,7 @@ "refId": "A" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (service_name) (rate(mongodb_mongod_op_latencies_latency_total{service_name=~\"$service_name\",type=\"command\"}[$interval]) / (rate(mongodb_mongod_op_latencies_ops_total{service_name=~\"$service_name\",type=\"command\"}[$interval]) > 0) or\nirate(mongodb_mongod_op_latencies_latency_total{service_name=~\"$service_name\",type=\"command\"}[5m]) / (irate(mongodb_mongod_op_latencies_ops_total{service_name=~\"$service_name\",type=\"command\"}[5m]) > 0))", @@ -2346,6 +2418,7 @@ "refId": "B" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (service_name) (max_over_time(mongodb_mongod_connections{service_name=~\"$service_name\", state=\"current\"}[$interval]) or \nmax_over_time(mongodb_mongod_connections{service_name=~\"$service_name\", state=\"current\"}[5m]) or\nmax_over_time(mongodb_mongos_connections{service_name=~\"$service_name\", state=\"current\"}[$interval]) or\nmax_over_time(mongodb_mongos_connections{service_name=~\"$service_name\", state=\"current\"}[5m]) or\nmax_over_time(mongodb_connections{service_name=~\"$service_name\", state=\"current\"}[$interval]) or\nmax_over_time(mongodb_connections{service_name=~\"$service_name\", state=\"current\"}[5m]))", @@ -2357,6 +2430,7 @@ "refId": "D" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (service_name) (max_over_time(mongodb_mongod_metrics_cursor_open{service_name=~\"$service_name\",state=\"total\"}[$interval]) or\nmax_over_time(mongodb_mongod_metrics_cursor_open{service_name=~\"$service_name\",state=\"total\"}[5m]) or\nmax_over_time(mongodb_mongod_cursors{service_name=~\"$service_name\",state=\"total\"}[$interval]) or\nmax_over_time(mongodb_mongod_cursors{service_name=~\"$service_name\",state=\"total\"}[5m]) or\nmax_over_time(mongodb_mongos_metrics_cursor_open{service_name=~\"$service_name\",state=\"total\"}[$interval]) or \nmax_over_time(mongodb_mongos_metrics_cursor_open{service_name=~\"$service_name\",state=\"total\"}[5m]) or\nmax_over_time(mongodb_mongos_cursors{service_name=~\"$service_name\",state=\"total\"}[$interval]) or\nmax_over_time(mongodb_mongos_cursors{service_name=~\"$service_name\",state=\"total\"}[5m]))", @@ -2367,16 +2441,56 @@ "refId": "E" } ], - "sorting_props": { - "col_index": -1, - "direction": "desc" - } + "title": "Services Details", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "agent_id": true, + "agent_type": true, + "environment": true, + "instance": true, + "job": true, + "machine_id": true, + "node_id": true, + "node_name": true, + "node_type": true, + "service_name": false + }, + "indexByName": { + "Time": 0, + "Value #A": 6, + "Value #B": 4, + "Value #C": 3, + "Value #D": 5, + "Value #E": 2, + "service_name": 1 + }, + "renameByName": { + "Value #A": "QPS", + "Value #B": "Average Latency", + "Value #C": "Uptime", + "Value #D": "Open Connections", + "Value #E": "Cursors", + "service_name": "Service Name" + } + } + } + ], + "type": "table" }, { "aliasColors": {}, "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Average latency of operations (classified by read, write, or (other) command).", "editable": true, @@ -2398,7 +2512,7 @@ "h": 8, "w": 24, "x": 0, - "y": 67 + "y": 107 }, "height": "250px", "hiddenSeries": false, @@ -2437,6 +2551,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name,type) (rate(mongodb_mongod_op_latencies_latency_total{service_name=~\"$service_name\"}[$interval]) / (rate(mongodb_mongod_op_latencies_ops_total{service_name=~\"$service_name\"}[$interval]) > 0) or irate(mongodb_mongod_op_latencies_latency_total{service_name=~\"$service_name\"}[5m]) / (irate(mongodb_mongod_op_latencies_ops_total{service_name=~\"$service_name\"}[5m]) > 0))", "format": "time_series", "hide": false, @@ -2492,6 +2607,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Docs per second inserted, updated, deleted or returned. (N.b. not 1-to-1 with operation counts.)", "editable": true, @@ -2513,7 +2629,7 @@ "h": 8, "w": 24, "x": 0, - "y": 75 + "y": 115 }, "height": "250px", "hiddenSeries": false, @@ -2552,6 +2668,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "avg by (service_name,state) (rate(mongodb_mongod_metrics_document_total{service_name=~\"$service_name\"}[$interval]) or \nirate(mongodb_mongod_metrics_document_total{service_name=~\"$service_name\"}[5m]))", "format": "time_series", "hide": false, @@ -2607,6 +2724,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "MongoDB keeps most recently used data in RAM. If you have created indexes for your queries and your working data set fits in RAM, MongoDB serves all queries from memory.", "editable": true, @@ -2628,7 +2746,7 @@ "h": 8, "w": 24, "x": 0, - "y": 83 + "y": 123 }, "height": "250px", "hiddenSeries": false, @@ -2667,6 +2785,7 @@ "steppedLine": false, "targets": [ { + "datasource": "Metrics", "expr": "rate(mongodb_tcmalloc_cache_bytes{service_name=~\"$service_name\"}[$interval]) or irate(mongodb_tcmalloc_cache_bytes{service_name=~\"$service_name\"}[5m])", "format": "time_series", "hide": false, @@ -2677,18 +2796,21 @@ "step": 300 }, { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_ss_tcmalloc_tcmalloc_thread_cache_free_bytes{service_name=~\"$service_name\"})", "interval": "$interval", "legendFormat": "Free Thread Cache", "refId": "A" }, { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_ss_tcmalloc_tcmalloc_central_cache_free_bytes{service_name=~\"$service_name\"})", "interval": "$interval", "legendFormat": "Free Central Cache", "refId": "B" }, { + "datasource": "Metrics", "expr": "avg by (service_name) (mongodb_ss_tcmalloc_tcmalloc_transfer_cache_free_bytes{service_name=~\"$service_name\"})", "interval": "$interval", "legendFormat": "Free Transfer Cache", @@ -2736,41 +2858,44 @@ } } ], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], "title": "MongoDB Services Summary", "type": "row" }, { "collapsed": true, + "datasource": "Metrics", "gridPos": { "h": 1, "w": 24, "x": 0, - "y": 59 + "y": 19 }, "id": 1403, "panels": [ { "collapsed": false, - "gridPos": { - "h": 6, - "w": 24, - "x": 0, - "y": 62 - }, - "id": 1435, - "pluginVersion": "9.2.20", + "datasource": "Metrics", "default_title_for_rows": "Node Name", - "title": "Nodes Overview", - "type": "table", "fieldConfig": { "defaults": { + "color": { + "fixedColor": "text", + "mode": "thresholds" + }, "custom": { "align": "center", "displayMode": "auto", - "inspect": false, "filterable": false, + "inspect": false, "width": 100 }, + "decimals": 2, "mappings": [ { "options": { @@ -2787,6 +2912,7 @@ "type": "value" } ], + "noValue": "-", "thresholds": { "mode": "absolute", "steps": [ @@ -2803,13 +2929,7 @@ "value": 20 } ] - }, - "color": { - "fixedColor": "text", - "mode": "thresholds" - }, - "decimals": 2, - "noValue": "-" + } }, "overrides": [ { @@ -2933,60 +3053,31 @@ } ] }, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 9 + }, + "id": 1435, "options": { - "showHeader": true, "footer": { + "fields": "", + "reducer": [ + "sum" + ], "show": false - } - }, - "transformations": [ - { - "id": "merge", - "options": {} }, - { - "id": "organize", - "options": { - "excludeByName": { - "Time": true, - "agent_id": true, - "agent_type": true, - "environment": true, - "instance": true, - "job": true, - "machine_id": true, - "node_id": true, - "node_name": false, - "node_type": true, - "service_name": false - }, - "indexByName": { - "Time": 0, - "Value #A": 3, - "Value #B": 5, - "Value #C": 6, - "Value #D": 2, - "Value #E": 8, - "Value #F": 4, - "Value #G": 7, - "node_name": 1 - }, - "renameByName": { - "Value #A": "Uptime", - "Value #B": "Load Average", - "Value #C": "Memory Available", - "Value #D": "RAM", - "Value #E": "Virtual Memory", - "Value #F": "Disk Space", - "Value #G": "Min Space Available", - "node_name": "Node Name", - "service_name": "Service Name" - } - } - } - ], + "showHeader": true + }, + "pluginVersion": "9.2.20", + "sorting_props": { + "col_index": -1, + "direction": "desc" + }, "targets": [ { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (node_name) (time() - container_start_time_seconds{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}) or avg by (node_name) ((node_time_seconds{node_name=~\"$node_name\"} - node_boot_time_seconds{node_name=~\"$node_name\"}) or (time() - node_boot_time_seconds{node_name=~\"$node_name\"}))", @@ -2998,6 +3089,7 @@ "refId": "A" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "max by (node_name) (avg_over_time(node_load1{node_name=~\"$node_name\"}[$interval]) or avg_over_time(node_load1{node_name=~\"$node_name\"}[5m]))", @@ -3009,6 +3101,7 @@ "refId": "B" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "clamp_max(avg by (node_name) (container_memory_usage_bytes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|backup-agent|\"} * 100 / container_spec_memory_limit_bytes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|backup-agent|\"}),100) or avg by (node_name) ((node_memory_MemAvailable_bytes{node_name=~\"$node_name\"} or (node_memory_MemFree_bytes{node_name=~\"$node_name\"} + node_memory_Buffers_bytes{node_name=~\"$node_name\"} + node_memory_Cached_bytes{node_name=~\"$node_name\"})) / node_memory_MemTotal_bytes{node_name=~\"$node_name\"} * 100)", @@ -3020,6 +3113,7 @@ "refId": "C" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "max by (node_name) (container_spec_memory_limit_bytes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}) or avg by (node_name) (node_memory_MemTotal_bytes{node_name=~\"$node_name\"})", @@ -3031,6 +3125,7 @@ "refId": "D" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "max by (node_name) (container_spec_memory_limit_bytes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}+container_spec_memory_swap_limit_bytes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}) or avg by (node_name) (node_memory_MemTotal_bytes{node_name=~\"$node_name\"}+node_memory_SwapTotal_bytes{node_name=~\"$node_name\"})", @@ -3041,6 +3136,7 @@ "refId": "E" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "avg by (node_name) (sum(avg(node_filesystem_size_bytes{node_name=~\"$node_name\",fstype=~\"(ext.|xfs|vfat|)\"}) without (mountpoint)) without (device,fstype))", @@ -3051,6 +3147,7 @@ "refId": "F" }, { + "datasource": "Metrics", "editorMode": "code", "exemplar": false, "expr": "min by (node_name) (node_filesystem_free_bytes{node_name=~\"$node_name\", fstype!~\"rootfs|selinuxfs|autofs|rpc_pipefs|tmpfs|shm|overlay|squashfs\"}/node_filesystem_size_bytes{node_name=~\"$node_name\", fstype!~\"rootfs|selinuxfs|autofs|rpc_pipefs|tmpfs|shm|overlay|squashfs\"})*100", @@ -3061,11 +3158,75 @@ "refId": "G" } ], - "sorting_props": { - "col_index": -1, - "direction": "desc" - } - }, + "title": "Nodes Overview", + "transformations": [ + { + "id": "merge", + "options": {} + }, + { + "id": "organize", + "options": { + "excludeByName": { + "Time": true, + "agent_id": true, + "agent_type": true, + "environment": true, + "instance": true, + "job": true, + "machine_id": true, + "node_id": true, + "node_name": false, + "node_type": true, + "service_name": false + }, + "indexByName": { + "Time": 0, + "Value #A": 3, + "Value #B": 5, + "Value #C": 6, + "Value #D": 2, + "Value #E": 8, + "Value #F": 4, + "Value #G": 7, + "node_name": 1 + }, + "renameByName": { + "Value #A": "Uptime", + "Value #B": "Load Average", + "Value #C": "Memory Available", + "Value #D": "RAM", + "Value #E": "Virtual Memory", + "Value #F": "Disk Space", + "Value #G": "Min Space Available", + "node_name": "Node Name", + "service_name": "Service Name" + } + } + } + ], + "type": "table" + } + ], + "targets": [ + { + "datasource": "Metrics", + "refId": "A" + } + ], + "title": "Nodes Summary", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 20 + }, + "id": 1569, + "panels": [ { "aliasColors": { "Max Core Utilization": "#bf1b00", @@ -3080,6 +3241,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "The CPU time is measured in clock ticks or seconds. It is useful to measure CPU time as a percentage of the CPU's capacity, which is called the CPU usage.", "editable": true, @@ -3095,9 +3257,9 @@ "grid": {}, "gridPos": { "h": 8, - "w": 24, + "w": 8, "x": 0, - "y": 68 + "y": 10 }, "height": "", "hiddenSeries": false, @@ -3126,7 +3288,7 @@ }, "paceLength": 10, "percentage": false, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "pointradius": 5, "points": false, "renderer": "flot", @@ -3146,6 +3308,7 @@ "steppedLine": true, "targets": [ { + "datasource": "Metrics", "expr": "avg by (node_name,mode) (clamp_max(((avg by (mode) ( (clamp_max(rate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\", mode!=\"user\", mode!=\"system\"}[$interval]),1)) or (clamp_max(irate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\", mode!=\"user\", mode!=\"system\"}[5m]),1)) ))*100 or (avg_over_time(node_cpu_average{node_name=~\"$node_name\", mode!=\"total\", mode!=\"idle\", mode!=\"user\", mode!=\"system\"}[$interval]) or avg_over_time(node_cpu_average{node_name=~\"$node_name\", mode!=\"total\", mode!=\"idle\", mode!=\"user\", mode!=\"system\"}[5m]))),100))", "format": "time_series", "interval": "$interval", @@ -3154,6 +3317,7 @@ "refId": "B" }, { + "datasource": "Metrics", "expr": "clamp_max(sum(rate(container_cpu_usage_seconds_total{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval]) or rate(container_cpu_usage_seconds_total{node_name=\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[5m])) or\nmax by () (sum by (cpu) ( (clamp_max(rate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\",mode!=\"iowait\"}[$interval]),1)) or (clamp_max(irate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\",mode!=\"iowait\"}[5m]),1)) )),1)", "format": "time_series", "hide": true, @@ -3163,6 +3327,7 @@ "refId": "C" }, { + "datasource": "Metrics", "exemplar": true, "expr": "clamp_max(avg(rate(container_cpu_system_seconds_total{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval])) or \n((avg by () ((clamp_max(rate(node_cpu_seconds_total{node_name=~\"$node_name\",mode=\"system\"}[$interval]),1)) or (clamp_max(irate(node_cpu_seconds_total{node_name=~\"$node_name\",mode=\"system\"}[5m]),1)) ))*100 or \n(avg_over_time(node_cpu_average{node_name=~\"$node_name\", mode=\"system\"}[$interval]) or \navg_over_time(node_cpu_average{node_name=~\"$node_name\", mode=\"system\"}[5m]))),100)", "hide": false, @@ -3171,6 +3336,7 @@ "refId": "A" }, { + "datasource": "Metrics", "exemplar": true, "expr": "clamp_max(avg(rate(container_cpu_user_seconds_total{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval])) or \n((avg by () ((clamp_max(rate(node_cpu_seconds_total{node_name=~\"$node_name\",mode=\"user\"}[$interval]),1)) or (clamp_max(irate(node_cpu_seconds_total{node_name=~\"$node_name\",mode=\"user\"}[5m]),1)) ))*100 or \n(avg_over_time(node_cpu_average{node_name=~\"$node_name\", mode=\"user\"}[$interval]) or \navg_over_time(node_cpu_average{node_name=~\"$node_name\", mode=\"user\"}[5m]))),100)", "hide": false, @@ -3214,7 +3380,21 @@ "yaxis": { "align": false } - }, + } + ], + "title": "CPU Usage", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 1571, + "panels": [ { "aliasColors": { "Allocated": "#E0752D", @@ -3228,6 +3408,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "When a system is running with maximum CPU utilization, the transmitting and receiving threads must all share the available CPU. This will cause data to be queued more frequently to cope with the lack of CPU. CPU Saturation may be measured as the length of a wait queue, or the time spent waiting on the queue.", "editable": true, @@ -3243,9 +3424,9 @@ "grid": {}, "gridPos": { "h": 8, - "w": 24, + "w": 8, "x": 0, - "y": 76 + "y": 11 }, "hiddenSeries": false, "id": 339, @@ -3272,7 +3453,7 @@ }, "paceLength": 10, "percentage": false, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "pointradius": 5, "points": false, "renderer": "flot", @@ -3293,6 +3474,7 @@ "targets": [ { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "((sum(avg_over_time(container_processes{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval]))-1) / avg by () (machine_cpu_cores)) or\navg by () ((avg_over_time(node_procs_running{node_name=~\"$node_name\"}[$interval])-1) / scalar(count(node_cpu_seconds_total{mode=\"user\", node_name=~\"$node_name\"})) or (avg_over_time(node_procs_running{node_name=~\"$node_name\"}[5m])-1) / scalar(count(node_cpu_seconds_total{mode=\"user\", node_name=~\"$node_name\"})))", @@ -3308,6 +3490,7 @@ }, { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "clamp_max(sum(rate(container_cpu_usage_seconds_total{node_name=~\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval]) or rate(container_cpu_usage_seconds_total{node_name=\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[5m])) or\nmax by () (sum by (cpu) ( (clamp_max(rate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\",mode!=\"iowait\"}[$interval]),1)) or (clamp_max(irate(node_cpu_seconds_total{node_name=~\"$node_name\",mode!=\"idle\",mode!=\"iowait\"}[5m]),1)) )),1)", @@ -3356,7 +3539,21 @@ "yaxis": { "align": false } - }, + } + ], + "title": "CPU Saturation", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 22 + }, + "id": 1585, + "panels": [ { "aliasColors": { "Swap In (Reads)": "#6ed0e0", @@ -3366,6 +3563,7 @@ "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Disk I/O includes read or write or input/output operations involving a physical disk. It is the speed with which the data transfer takes place between the hard disk drive and RAM.\n\nSwap Activity is memory management that involves swapping sections of memory to and from physical storage.", "editable": true, @@ -3381,9 +3579,9 @@ "grid": {}, "gridPos": { "h": 8, - "w": 24, + "w": 8, "x": 0, - "y": 84 + "y": 35 }, "hiddenSeries": false, "id": 341, @@ -3415,7 +3613,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "pointradius": 5, "points": false, "renderer": "flot", @@ -3442,6 +3640,7 @@ "targets": [ { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "avg by () (sum(rate(container_fs_reads_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_reads_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) or\navg by () (rate(node_vmstat_pgpgin{node_name=\"$node_name\"}[$interval]) * 1024 or irate(node_vmstat_pgpgin{node_name=\"$node_name\"}[5m]) * 1024)", @@ -3456,6 +3655,7 @@ }, { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "avg by () (sum(rate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) or\navg by () ((rate(node_vmstat_pgpgout{node_name=\"$node_name\"}[$interval]) * 1024 or irate(node_vmstat_pgpgout{node_name=\"$node_name\"}[5m]) * 1024))", @@ -3469,6 +3669,7 @@ "target": "" }, { + "datasource": "Metrics", "expr": "avg by () (sum(rate(container_fs_reads_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_reads_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) + \navg by () (sum(rate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) or \navg by () ((rate(node_vmstat_pgpgin{node_name=\"$node_name\"}[$interval]) * 1024 or irate(node_vmstat_pgpgin{node_name=\"$node_name\"}[5m]) * 1024 ) + (rate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) + \navg by () (sum(rate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]) or irate(container_fs_writes_bytes_total{node_name=\"$node_name\",id=~\"/kubepods.*\"}[5s]))) or \navg by () ((rate(node_vmstat_pgpgout{node_name=\"$node_name\"}[$interval]) * 1024 or irate(node_vmstat_pgpgout{node_name=\"$node_name\"}[5m]) * 1024))", "format": "time_series", "interval": "$interval", @@ -3477,6 +3678,7 @@ "refId": "C" }, { + "datasource": "Metrics", "expr": "avg by (node_name) (rate(node_vmstat_pswpin{node_name=\"$node_name\"}[$interval]) * 4096 or irate(node_vmstat_pswpin{node_name=\"$node_name\"}[5m]) * 4096)", "format": "time_series", "interval": "$interval", @@ -3485,6 +3687,7 @@ "refId": "D" }, { + "datasource": "Metrics", "expr": "avg by () (rate(container_memory_swap{node_name=\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[$interval]) or \nirate(container_memory_swap{node_name=\"$node_name\",id=~\"/kubepods.*\",container!~\"POD|pmm-client|\"}[5s])) or \navg by () (rate(node_vmstat_pswpout{node_name=\"$node_name\"}[$interval]) * 4096 or \nirate(node_vmstat_pswpout{node_name=\"$node_name\"}[5m]) * 4096)", "format": "time_series", "interval": "$interval", @@ -3526,12 +3729,27 @@ "yaxis": { "align": false } - }, + } + ], + "title": "Disk I/O and Swap Activity", + "type": "row" + }, + { + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 23 + }, + "id": 1583, + "panels": [ { "aliasColors": {}, "bars": false, "dashLength": 10, "dashes": false, + "datasource": "Metrics", "decimals": 2, "description": "Network traffic refers to the amount of data moving across a network at a given point in time.", "editable": true, @@ -3547,9 +3765,9 @@ "grid": {}, "gridPos": { "h": 8, - "w": 24, + "w": 8, "x": 0, - "y": 92 + "y": 36 }, "hiddenSeries": false, "id": 343, @@ -3575,7 +3793,7 @@ "alertThreshold": true }, "percentage": false, - "pluginVersion": "8.3.5", + "pluginVersion": "9.2.20", "pointradius": 5, "points": false, "renderer": "flot", @@ -3593,6 +3811,7 @@ "targets": [ { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "sum(rate(container_network_receive_bytes_total{node_name=\"$node_name\",id=~\"/kubepods/.*\"}[$interval]) or \nirate(container_network_receive_bytes_total{node_name=\"$node_name\",id=~\"/kubepods/.*\"}[5m])) or\nsum(rate(node_network_receive_bytes_total{node_name=\"$node_name\", device!=\"lo\"}[$interval])) or \nsum(irate(node_network_receive_bytes_total{node_name=\"$node_name\", device!=\"lo\"}[5m])) or \nsum(max_over_time(rdsosmetrics_network_rx{node_name=\"$node_name\"}[$interval])) or \nsum(max_over_time(rdsosmetrics_network_rx{node_name=\"$node_name\"}[5m]))", @@ -3607,6 +3826,7 @@ }, { "calculatedInterval": "2s", + "datasource": "Metrics", "datasourceErrors": {}, "errors": {}, "expr": "sum(rate(container_network_transmit_bytes_total{node_name=\"$node_name\",id=~\"/kubepods/.*\"}[$interval]) or \nirate(container_network_transmit_bytes_total{node_name=\"$node_name\",id=~\"/kubepods/.*\"}[5m])) or \nsum(rate(node_network_transmit_bytes_total{node_name=\"$node_name\", device!=\"lo\"}[$interval])) or \nsum(irate(node_network_transmit_bytes_total{node_name=\"$node_name\", device!=\"lo\"}[5m])) or\nsum(max_over_time(rdsosmetrics_network_tx{node_name=\"$node_name\"}[$interval])) or \nsum(max_over_time(rdsosmetrics_network_tx{node_name=\"$node_name\"}[5m]))", @@ -3655,12 +3875,12 @@ } } ], - "title": "Nodes Summary", + "title": "Network Traffic", "type": "row" } ], "refresh": "1m", - "schemaVersion": 34, + "schemaVersion": 37, "style": "dark", "tags": [ "MongoDB_HA", @@ -3736,8 +3956,8 @@ "allFormat": "blob", "current": { "selected": false, - "text": "", - "value": "" + "text": "igtest", + "value": "igtest" }, "datasource": "Metrics", "definition": "label_values(mongodb_up,cluster)", @@ -3763,8 +3983,8 @@ "allFormat": "glob", "current": { "selected": false, - "text": "", - "value": "" + "text": "shard0", + "value": "shard0" }, "datasource": "Metrics", "definition": "label_values(mongodb_mongod_replset_my_state{cluster=~\"$cluster\"}, set)", @@ -3789,16 +4009,20 @@ { "allFormat": "glob", "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "All" + ], + "value": [ + "$__all" + ] }, "datasource": "Metrics", "definition": "label_values(mongodb_mongod_replset_my_state{service_name=~\"$service_name\"}, node_name)", - "hide": 2, - "includeAll": false, + "hide": 0, + "includeAll": true, "label": "Node Name", - "multi": false, + "multi": true, "multiFormat": "glob", "name": "node_name", "options": [], @@ -3817,9 +4041,13 @@ { "allFormat": "glob", "current": { - "selected": false, - "text": "All", - "value": "$__all" + "selected": true, + "text": [ + "All" + ], + "value": [ + "$__all" + ] }, "datasource": "Metrics", "definition": "",