Add similar changes to ovn dashboard, remove old queries

cloud-bulldozer · Sep 25, 2024 · ffb3daf · ffb3daf
1 parent 75952e3
commit ffb3daf
Show file tree

Hide file tree

Showing 4 changed files with 16 additions and 29 deletions.
diff --git a/assets/ocp-performance/queries.libsonnet b/assets/ocp-performance/queries.libsonnet
@@ -150,19 +150,19 @@ local generateTimeSeriesQuery(query, legend) = [
   },
   ovnKubeControlPlaneCPU: {
     query():
-      generateTimeSeriesQuery('irate(container_cpu_usage_seconds_total{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}[$interval])*100', '{{container}}-{{pod}}-{{node}}'),
+      generateTimeSeriesQuery('sum( irate(container_cpu_usage_seconds_total{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}[$interval])*100 ) by (pod, node)', '{{pod}} - {{node}}'),
   },
   ovnKubeControlPlaneMem: {
     query():
-      generateTimeSeriesQuery('container_memory_rss{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}', '{{container}}-{{pod}}-{{node}}'),
+      generateTimeSeriesQuery('container_memory_rss{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}', '{{pod}} - {{node}}'),
   },
   topOvnControllerCPU: {
     query():
-      generateTimeSeriesQuery('topk(10, irate(container_cpu_usage_seconds_total{pod=~"ovnkube-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}[$interval])*100)', '{{node}}'),
+      generateTimeSeriesQuery('topk(10, sum( irate(container_cpu_usage_seconds_total{pod=~"ovnkube-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}[$interval])*100)  by (pod,node) )', '{{pod}} - {{node}}'),
   },
   topOvnControllerMem: {
     query():
-      generateTimeSeriesQuery('topk(10, sum(container_memory_rss{pod=~"ovnkube-node-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}) by (node))', '{{node}}'),
+      generateTimeSeriesQuery('topk(10, sum(container_memory_rss{pod=~"ovnkube-node-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}) by (pod,node))', '{{pod}} - {{node}}'),
   },
   promReplCpuUsage: {
     query():

diff --git a/assets/ovn-monitoring/queries.libsonnet b/assets/ovn-monitoring/queries.libsonnet
@@ -10,49 +10,38 @@ local generateTimeSeriesQuery(query, legend) = [
 ];
 
 {
-  ovnMasterLeader: {
+  ovnClusterManagerLeader: {
     query():
-      generateTimeSeriesQuery('ovnkube_master_leader', '{{pod}}'),
+      generateTimeSeriesQuery('ovnkube_clustermanager_leader > 0', '{{pod}}'),
   },
 
   ovnNorthd: {
     query():
       generateTimeSeriesQuery('ovn_northd_status', '{{pod}}'),
   },
 
-  ovnNbdbLeader: {
-    query():
-      generateTimeSeriesQuery('ovn_db_cluster_server_role{server_role="leader",db_name="OVN_Northbound"}', '{{pod}}'),
-  },
-
-  ovnSbdbLeader: {
-    query():
-      generateTimeSeriesQuery('ovn_db_cluster_server_role{server_role="leader",db_name="OVN_Southbound"}', '{{pod}}'),
-  },
-
   numOnvController: {
     query():
       generateTimeSeriesQuery('count(ovn_controller_monitor_all) by (namespace)', ''),
   },
 
   ovnKubeControlPlaneCPU: {
     query():
-      generateTimeSeriesQuery('irate(container_cpu_usage_seconds_total{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}[2m])*100','{{container}}-{{pod}}-{{node}}'),
+      generateTimeSeriesQuery('sum( irate(container_cpu_usage_seconds_total{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}[2m])*100 ) by (pod, node)', '{{pod}} - {{node}}'),
   },
 
   ovnKubeControlPlaneMem: {
     query():
-      generateTimeSeriesQuery('container_memory_rss{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}','{{container}}-{{pod}}-{{node}}'),
+      generateTimeSeriesQuery('container_memory_rss{pod=~"(ovnkube-master|ovnkube-control-plane).+",namespace="openshift-ovn-kubernetes",container!~"POD|"}', '{{pod}} - {{node}}'),
   },
 
   topOvnControllerCPU: {
     query():
-      generateTimeSeriesQuery('topk(10, irate(container_cpu_usage_seconds_total{pod=~"ovnkube-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}[2m])*100)', '{{node}}'),
+      generateTimeSeriesQuery('topk(10, sum( irate(container_cpu_usage_seconds_total{pod=~"ovnkube-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}[2m])*100)  by (pod,node) )', '{{pod}} - {{node}}'),
   },
-
   topOvnControllerMem: {
     query():
-      generateTimeSeriesQuery('topk(10, sum(container_memory_rss{pod=~"ovnkube-node-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}) by (node))', '{{node}}'),
+      generateTimeSeriesQuery('topk(10, sum(container_memory_rss{pod=~"ovnkube-node-.*",namespace="openshift-ovn-kubernetes",container="ovn-controller"}) by (pod,node))', '{{pod}} - {{node}}'),
   },
 
   ovnAnnotationLatency: {

diff --git a/templates/General/ocp-performance.jsonnet b/templates/General/ocp-performance.jsonnet
@@ -55,9 +55,9 @@ g.dashboard.new('Openshift Performance')
     panels.timeSeries.genericLegend('ovs-master Memory Usage', 'bytes', queries.OVSMemory.query('$_master_node'), { x: 12, y: 17, w: 12, h: 8 }),
     panels.timeSeries.genericLegend('ovs-worker CPU Usage', 'percent', queries.OVSCPU.query('$_worker_node'), { x: 0, y: 25, w: 12, h: 8 }),
     panels.timeSeries.genericLegend('ovs-worker Memory Usage', 'bytes', queries.OVSMemory.query('$_worker_node'), { x: 12, y: 25, w: 12, h: 8 }),
-    panels.timeSeries.genericLegend('99% Pod Annotation Latency', 's', queries.ovnAnnotationLatency.query(), { x: 0, y: 33, w: 24, h: 8 }),
-    panels.timeSeries.genericLegend('99% CNI Request ADD Latency', 's', queries.ovnCNIAdd.query(), { x: 0, y: 41, w: 12, h: 8 }),
-    panels.timeSeries.genericLegend('99% CNI Request DEL Latency', 's', queries.ovnCNIDel.query(), { x: 12, y: 41, w: 12, h: 8 }),
+    panels.timeSeries.genericLegend('99% Pod Annotation Latency', 's', queries.ovnAnnotationLatency.query(), { x: 0, y: 33, w: 8, h: 8 }),
+    panels.timeSeries.genericLegend('99% CNI Request ADD Latency', 's', queries.ovnCNIAdd.query(), { x: 8, y: 41, w: 8, h: 8 }),
+    panels.timeSeries.genericLegend('99% CNI Request DEL Latency', 's', queries.ovnCNIDel.query(), { x: 16, y: 41, w: 8, h: 8 }),
   ]),
   g.panel.row.new('Monitoring stack')
   + g.panel.row.withGridPos({ x: 0, y: 0, w: 24, h: 1 })

diff --git a/templates/General/ovn-dashboard.jsonnet b/templates/General/ovn-dashboard.jsonnet
@@ -26,11 +26,9 @@ g.dashboard.new('OVN-Monitoring-dashboard')
   + g.panel.row.withCollapsed(true)
   + g.panel.row.withGridPos({ x: 0, y: 0, w: 24, h: 1 })
   + g.panel.row.withPanels([
-    panels.stat.genericstatThresoldPanel('OVNKube Master', 'none', queries.ovnMasterLeader.query(), { x: 0, y: 0, w: 4, h: 4 }),
-    panels.stat.genericstatThresoldPanel('OVN Northd Status', 'none', queries.ovnNorthd.query(), { x: 4, y: 0, w: 4, h: 4 }),
-    panels.stat.genericstatThresoldPanel('OVN NBDB leader', 'none', queries.ovnNbdbLeader.query(), { x: 8, y: 0, w: 4, h: 4 }),
-    panels.stat.genericstatThresoldPanel('OVN SBDB leader', 'none', queries.ovnSbdbLeader.query(), { x: 12, y: 0, w: 4, h: 4 }),
-    panels.stat.genericstatThresoldOVNControllerPanel('OVN controller', 'none', queries.numOnvController.query(), { x: 16, y: 0, w: 4, h: 4 }),
+    panels.stat.genericstatThresoldPanel('OVNKube Master', 'none', queries.ovnClusterManagerLeader.query(), { x: 0, y: 0, w: 8, h: 4 }),
+    panels.stat.genericstatThresoldPanel('OVN Northd Status', 'none', queries.ovnNorthd.query(), { x: 8, y: 0, w: 8, h: 4 }),
+    panels.stat.genericstatThresoldOVNControllerPanel('OVN controller', 'none', queries.numOnvController.query(), { x: 16, y: 0, w: 8, h: 4 }),
     panels.timeSeries.genericTimeSeriesLegendPanel('ovnkube-control-plane CPU Usage', 'percent', queries.ovnKubeControlPlaneCPU.query(), { x: 0, y: 4, w: 12, h: 10 }),
     panels.timeSeries.genericTimeSeriesLegendPanel('ovnkube-control-plane Memory Usage', 'bytes', queries.ovnKubeControlPlaneMem.query(), { x: 12, y: 4, w: 12, h: 10 }),
     panels.timeSeries.genericTimeSeriesLegendPanel('Top 10 ovn-controller CPU Usage', 'percent', queries.topOvnControllerCPU.query(), { x: 0, y: 12, w: 12, h: 10 }),