Add cgroup RSS queries

Signed-off-by: Andrew Collins <[email protected]>
kube-burner · Nov 13, 2024 · bbffbc4 · bbffbc4
1 parent ef7986d
commit bbffbc4
Show file tree

Hide file tree

Showing 4 changed files with 85 additions and 0 deletions.
diff --git a/cmd/config/metrics-aggregated.yml b/cmd/config/metrics-aggregated.yml
@@ -156,17 +156,37 @@
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "worker",role != "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Workers
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "master" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Masters
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "master" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Masters
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Infra
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Infra
+  instant: true
+  captureStart: true
+
 - query: sum( container_cpu_usage_seconds_total{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
   metricName: cgroupCPUSeconds-namespaces
   instant: true
   captureStart: true
+
+- query: sum( container_memory_rss{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
+  metricName: cgroupMemoryRSS-namespaces
+  instant: true
+  captureStart: true
diff --git a/cmd/config/metrics-egressip.yml b/cmd/config/metrics-egressip.yml
@@ -168,17 +168,37 @@
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "worker",role != "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Workers
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "master" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Masters
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "master" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Masters
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Infra
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Infra
+  instant: true
+  captureStart: true
+
 - query: sum( container_cpu_usage_seconds_total{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
   metricName: cgroupCPUSeconds-namespaces
   instant: true
   captureStart: true
+
+- query: sum( container_memory_rss{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
+  metricName: cgroupMemoryRSS-namespaces
+  instant: true
+  captureStart: true
diff --git a/cmd/config/metrics-report.yml b/cmd/config/metrics-report.yml
@@ -300,6 +300,11 @@
   metricName: max-memory-sum-workers
   instant: true
 
+- query: sum( (node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) and on (instance) label_replace(kube_node_role{role="worker"}, "instance", "$1", "node", "(.+)") )
+  metricName: memory-sum-workers
+  instant: true
+  captureStart: true
+
 - query: avg(avg_over_time(sum(irate(node_cpu_seconds_total{mode!="idle", mode!="steal"}[2m]) and on (instance) label_replace(kube_node_role{role="infra"}, "instance", "$1", "node", "(.+)")) by (instance)[{{.elapsed}}:]))
   metricName: cpu-infra
   instant: true
@@ -393,17 +398,37 @@
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "worker",role != "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Workers
+  instant: true
+  captureStart: true
+
 - query: sum(container_cpu_usage_seconds_total{id=~"/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice"} and on (node) kube_node_role{role="master"}) by (id)
   metricName: cgroupCPUSeconds-Masters
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "master" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Masters
+  instant: true
+  captureStart: true
+
 - query: sum(container_cpu_usage_seconds_total{id=~"/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice"} and on (node) kube_node_role{role="infra"}) by (id)
   metricName: cgroupCPUSeconds-Infra
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Infra
+  instant: true
+  captureStart: true
+
 - query: sum(container_cpu_usage_seconds_total{container!~"POD|",namespace=~"openshift-.*"}) by (namespace)
   metricName: cgroupCPUSeconds-namespaces
   instant: true
   captureStart: true
+
+- query: sum( container_memory_rss{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
+  metricName: cgroupMemoryRSS-namespaces
+  instant: true
+  captureStart: true
diff --git a/cmd/config/metrics.yml b/cmd/config/metrics.yml
@@ -147,17 +147,37 @@
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "worker",role != "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Workers
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "master" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Masters
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "master" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Masters
+  instant: true
+  captureStart: true
+
 - query: sum (  container_cpu_usage_seconds_total {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" }  and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
   metricName: cgroupCPUSeconds-Infra
   instant: true
   captureStart: true
 
+- query: sum (  container_memory_rss {  id =~ "/system.slice|/system.slice/kubelet.service|/system.slice/ovs-vswitchd.service|/system.slice/crio.service|/kubepods.slice" } and on (node) kube_node_role{ role = "infra" } )  by   (   id  )
+  metricName: cgroupMemoryRSS-Infra
+  instant: true
+  captureStart: true
+
 - query: sum( container_cpu_usage_seconds_total{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
   metricName: cgroupCPUSeconds-namespaces
   instant: true
   captureStart: true
+
+- query: sum( container_memory_rss{container!~"POD|",namespace=~"openshift-.*"} )  by (namespace)
+  metricName: cgroupMemoryRSS-namespaces
+  instant: true
+  captureStart: true