add inclusternetworklatency variable and arm64 toleration

mm4tt · Jul 21, 2022 · f7c1fd7 · f7c1fd7
1 parent f726928
commit f7c1fd7
Show file tree

Hide file tree

Showing 7 changed files with 45 additions and 0 deletions.
diff --git a/clusterloader2/testing/huge-service/modules/measurements.yaml b/clusterloader2/testing/huge-service/modules/measurements.yaml
@@ -8,6 +8,7 @@
 {{$ENABLE_VIOLATIONS_FOR_API_CALL_PROMETHEUS_SIMPLE := DefaultParam .CL2_ENABLE_VIOLATIONS_FOR_API_CALL_PROMETHEUS_SIMPLE true}}
 
 {{$CUSTOM_API_CALL_THRESHOLDS := DefaultParam .CUSTOM_API_CALL_THRESHOLDS ""}}
+{{$ENABLE_IN_CLUSTER_NETWORK_LATENCY := DefaultParam .CL2_ENABLE_IN_CLUSTER_NETWORK_LATENCY true}}
 {{$ENABLE_RESTART_COUNT_CHECK := DefaultParam .ENABLE_RESTART_COUNT_CHECK true}}
 {{$ENABLE_SYSTEM_POD_METRICS := DefaultParam .ENABLE_SYSTEM_POD_METRICS true}}
 {{$RESTART_COUNT_THRESHOLD_OVERRIDES := DefaultParam .RESTART_COUNT_THRESHOLD_OVERRIDES ""}}
@@ -50,12 +51,14 @@ steps:
       enableRestartCountCheck: {{$ENABLE_RESTART_COUNT_CHECK}}
       clusterOOMsIgnoredProcesses: {{YamlQuote $CLUSTER_OOMS_IGNORED_PROCESSES 4}}
       clusterOOMsTrackerEnabled: {{$ENABLE_CLUSTER_OOMS_TRACKER}}
+{{if $ENABLE_IN_CLUSTER_NETWORK_LATENCY}}
   - Identifier: InClusterNetworkLatency
     Method: InClusterNetworkLatency
     Params:
       action: {{$action}}
       checkProbesReadyTimeout: {{$PROBE_MEASUREMENTS_CHECK_PROBES_READY_TIMEOUT}}
       replicasPerProbe: {{AddInt 2 (DivideInt .Nodes 100)}}
+{{end}}
 {{if $ENABLE_CONTAINER_RESTARTS_MEASUREMENT}}
   - Identifier: ContainerRestarts
     Method: ContainerRestarts

diff --git a/clusterloader2/testing/load/daemonset.yaml b/clusterloader2/testing/load/daemonset.yaml
@@ -2,6 +2,7 @@
 {{$Image := DefaultParam .Image "k8s.gcr.io/pause:3.1"}}
 {{$Env := DefaultParam .Env ""}}
 {{$DaemonSetSurge := DefaultParam .CL2_DS_SURGE (MaxInt 10 (DivideInt .Nodes 20))}} # 5% of nodes, but not less than 10
+{{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
 
 apiVersion: apps/v1
 kind: DaemonSet
@@ -37,3 +38,10 @@ spec:
             memory: "20M"
       priorityClassName: daemonset-priorityclass-0 # Name is autogenerated, hence the -0 prefix.
       terminationGracePeriodSeconds: 1
+      tolerations: 
+      {{if $RUN_ON_ARM_NODES}}
+      - key: "kubernetes.io/arch"
+        operator: Equal
+        value: arm64
+        effect: NoSchedule
+      {{end}}
diff --git a/clusterloader2/testing/load/deployment.yaml b/clusterloader2/testing/load/deployment.yaml
@@ -6,6 +6,7 @@
 {{$dnsQPSPerClient := DefaultParam .CL2_DNS_QPS_PER_CLIENT 1}}
 # Guard the new DNS tests. Remove it once it's confirmed that it works on a subset of tests.
 {{$USE_ADVANCED_DNSTEST := DefaultParam .CL2_USE_ADVANCED_DNSTEST false}}
+{{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
 
 apiVersion: apps/v1
 kind: Deployment
@@ -78,6 +79,12 @@ spec:
         operator: "Exists"
         effect: "NoExecute"
         tolerationSeconds: 900
+      {{if $RUN_ON_ARM_NODES}}
+      - key: "kubernetes.io/arch"
+        operator: Equal
+        value: arm64
+        effect: NoSchedule
+      {{end}}
       volumes:
         - name: configmap
           configMap:

diff --git a/clusterloader2/testing/load/job.yaml b/clusterloader2/testing/load/job.yaml
@@ -1,4 +1,5 @@
 {{$HostNetworkMode := DefaultParam .CL2_USE_HOST_NETWORK_PODS false}}
+{{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
 
 apiVersion: batch/v1
 kind: Job
@@ -43,3 +44,9 @@ spec:
           operator: "Exists"
           effect: "NoExecute"
           tolerationSeconds: 900
+        {{if $RUN_ON_ARM_NODES}}
+        - key: "kubernetes.io/arch"
+          operator: Equal
+          value: arm64
+          effect: NoSchedule
+        {{end}}
diff --git a/clusterloader2/testing/load/modules/measurements.yaml b/clusterloader2/testing/load/modules/measurements.yaml
@@ -10,6 +10,8 @@
 {{$CLUSTER_OOMS_IGNORED_PROCESSES := DefaultParam .CL2_CLUSTER_OOMS_IGNORED_PROCESSES ""}}
 {{$CUSTOM_API_CALL_THRESHOLDS := DefaultParam .CUSTOM_API_CALL_THRESHOLDS ""}}
 {{$ENABLE_API_AVAILABILITY_MEASUREMENT := DefaultParam .CL2_ENABLE_API_AVAILABILITY_MEASUREMENT false}}
+{{$ENABLE_IN_CLUSTER_NETWORK_LATENCY := DefaultParam .CL2_ENABLE_IN_CLUSTER_NETWORK_LATENCY true}}
+{{$ENABLE_SLO_MEASUREMENT := DefaultParam .CL2_ENABLE_SLO_MEASUREMENT true}}
 {{$ENABLE_CLUSTER_OOMS_TRACKER := DefaultParam .CL2_ENABLE_CLUSTER_OOMS_TRACKER true}}
 {{$ENABLE_DNSTESTS := DefaultParam .CL2_ENABLE_DNSTESTS false}}
 # Guard the new DNS tests. Remove it once it's confirmed that it works on a subset of tests.
@@ -64,12 +66,14 @@ steps:
       action: {{$action}}
       labelSelector: group = load
       threshold: 1h # TODO(https://github.com/kubernetes/perf-tests/issues/1024): Ideally, this should be 5s
+{{if $ENABLE_IN_CLUSTER_NETWORK_LATENCY}}
   - Identifier: InClusterNetworkLatency
     Method: InClusterNetworkLatency
     Params:
       action: {{$action}}
       checkProbesReadyTimeout: {{$PROBE_MEASUREMENTS_CHECK_PROBES_READY_TIMEOUT}}
       replicasPerProbe: {{AddInt 2 (DivideInt .Nodes 100)}}
+{{end}}
 {{if $ENABLE_NODE_LOCAL_DNS_LATENCY}}
   - Identifier: NodeLocalDNSLatency
     Method: NodeLocalDNSLatencyPrometheus
@@ -78,12 +82,14 @@ steps:
       enableViolations: true
       threshold: {{$NODE_LOCAL_DNS_LATENCY_THRESHOLD}}
 {{end}}
+{{if $ENABLE_SLO_MEASUREMENT}}
   - Identifier: SLOMeasurement
     Method: SLOMeasurement
     Params:
       action: {{$action}}
       checkProbesReadyTimeout: {{$PROBE_MEASUREMENTS_CHECK_PROBES_READY_TIMEOUT}}
       replicasPerProbe: {{AddInt 2 (DivideInt .Nodes 100)}}
+{{end}}
 {{if $PROMETHEUS_SCRAPE_KUBE_PROXY}}
   - Identifier: NetworkProgrammingLatency
     Method: NetworkProgrammingLatency

diff --git a/clusterloader2/testing/load/simple-deployment.yaml b/clusterloader2/testing/load/simple-deployment.yaml
@@ -4,6 +4,7 @@
 {{$CpuRequest := DefaultParam .CpuRequest "5m"}}
 {{$MemoryRequest := DefaultParam .MemoryRequest "20M"}}
 {{$Image := DefaultParam .Image "k8s.gcr.io/pause:3.1"}}
+{{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
 
 apiVersion: apps/v1
 kind: Deployment
@@ -46,3 +47,9 @@ spec:
         operator: "Exists"
         effect: "NoExecute"
         tolerationSeconds: 900
+      {{if $RUN_ON_ARM_NODES}}
+      - key: "kubernetes.io/arch"
+        operator: Equal
+        value: arm64
+        effect: NoSchedule
+      {{end}}
diff --git a/clusterloader2/testing/load/statefulset.yaml b/clusterloader2/testing/load/statefulset.yaml
@@ -1,5 +1,6 @@
 {{$HostNetworkMode := DefaultParam .CL2_USE_HOST_NETWORK_PODS false}}
 {{$EnablePVs := DefaultParam .CL2_ENABLE_PVS true}}
+{{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
 
 apiVersion: apps/v1
 kind: StatefulSet
@@ -51,6 +52,12 @@ spec:
         operator: "Exists"
         effect: "NoExecute"
         tolerationSeconds: 900
+      {{if $RUN_ON_ARM_NODES}}
+      - key: "kubernetes.io/arch"
+        operator: Equal
+        value: arm64
+        effect: NoSchedule
+      {{end}}
   {{if $EnablePVs}}
   # NOTE: PVs created this way should be cleaned-up manually, as deleting the StatefulSet doesn't automatically delete PVs.
   # To avoid deleting all the PVs at once during namespace deletion, they should be deleted explicitly via Phase.