clusterloader2: add CL2_TOLERATION override to all workload templates

Add a new CL2_TOLERATION parameter that, when set to a non-empty string,
adds a toleration with operator: Exists and effect: NoSchedule using that
string as the key. This allows directing all workloads to any arbitrarily
tainted node without needing a dedicated override per taint type.

Applies to: Deployment, StatefulSet, DaemonSet, Job (reconcile-objects)
and simple-deployment (scheduler-throughput / huge-service).

Co-authored-by: Lei Yao <[email protected]>

Author: wonderyl

clusterloader2/testing/load/modules/reconcile-objects/daemonset.yaml

@@ -3,6 +3,7 @@
 {{$Env := DefaultParam .Env ""}}
 {{$DaemonSetSurge := DefaultParam .CL2_DS_SURGE (MaxInt 10 (DivideInt .Nodes 20))}} # 5% of nodes, but not less than 10
 {{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
+{{$TOLERATION := DefaultParam .CL2_TOLERATION ""}}
 
 {{$ENABLE_NETWORK_POLICY_ENFORCEMENT_LATENCY_TEST := DefaultParam .CL2_ENABLE_NETWORK_POLICY_ENFORCEMENT_LATENCY_TEST false}}
 {{$NET_POLICY_ENFORCEMENT_LATENCY_NODE_LABEL_VALUE := DefaultParam .CL2_NET_POLICY_ENFORCEMENT_LATENCY_NODE_LABEL_VALUE "net-policy-client"}}
@@ -55,6 +56,11 @@ spec:
         value: arm64
         effect: NoSchedule
       {{end}}
+      {{if $TOLERATION}}
+      - key: {{$TOLERATION}}
+        operator: Exists
+        effect: NoSchedule
+      {{end}}
       {{if $ENABLE_NETWORK_POLICY_ENFORCEMENT_LATENCY_TEST}}
       - key: test-np
         operator: Equal

clusterloader2/testing/load/modules/reconcile-objects/deployment.yaml

@@ -8,6 +8,7 @@
 # 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}}
+{{$TOLERATION := DefaultParam .CL2_TOLERATION ""}}
 
 {{$EnableNetworkPolicyEnforcementLatencyTest := DefaultParam .EnableNetworkPolicyEnforcementLatencyTest false}}
 {{$TargetLabelValue := DefaultParam .TargetLabelValue "enforcement-latency"}}
@@ -113,6 +114,11 @@ spec:
         value: arm64
         effect: NoSchedule
       {{end}}
+      {{if $TOLERATION}}
+      - key: {{$TOLERATION}}
+        operator: Exists
+        effect: NoSchedule
+      {{end}}
       volumes:
         - name: configmap
           configMap:

clusterloader2/testing/load/modules/reconcile-objects/job.yaml

@@ -1,5 +1,6 @@
 {{$HostNetworkMode := DefaultParam .CL2_USE_HOST_NETWORK_PODS false}}
 {{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
+{{$TOLERATION := DefaultParam .CL2_TOLERATION ""}}
 {{$Image := DefaultParam .Image "registry.k8s.io/pause:3.9"}}
 {{$podPayloadSize := DefaultParam .CL2_JOB_POD_PAYLOAD_SIZE 0}}
 {{$RuntimeClassName := DefaultParam .CL2_RUNTIME_CLASS_NAME nil}}
@@ -60,3 +61,8 @@ spec:
           value: arm64
           effect: NoSchedule
         {{end}}
+        {{if $TOLERATION}}
+        - key: {{$TOLERATION}}
+          operator: Exists
+          effect: NoSchedule
+        {{end}}

clusterloader2/testing/load/modules/reconcile-objects/statefulset.yaml

@@ -1,6 +1,7 @@
 {{$HostNetworkMode := DefaultParam .CL2_USE_HOST_NETWORK_PODS false}}
 {{$EnablePVs := DefaultParam .CL2_ENABLE_PVS true}}
 {{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
+{{$TOLERATION := DefaultParam .CL2_TOLERATION ""}}
 {{$podPayloadSize := DefaultParam .CL2_STATEFULSET_POD_PAYLOAD_SIZE 0}}
 {{$Image := DefaultParam .Image "registry.k8s.io/pause:3.9"}}
 {{$RuntimeClassName := DefaultParam .CL2_RUNTIME_CLASS_NAME nil}}
@@ -67,6 +68,11 @@ spec:
         value: arm64
         effect: NoSchedule
       {{end}}
+      {{if $TOLERATION}}
+      - key: {{$TOLERATION}}
+        operator: Exists
+        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.

clusterloader2/testing/load/modules/scheduler-throughput/simple-deployment.yaml

@@ -6,6 +6,7 @@
 {{$MemoryRequest := DefaultParam .MemoryRequest "20M"}}
 {{$Image := DefaultParam .Image "registry.k8s.io/pause:3.9"}}
 {{$RUN_ON_ARM_NODES := DefaultParam .CL2_RUN_ON_ARM_NODES false}}
+{{$TOLERATION := DefaultParam .CL2_TOLERATION ""}}
 
 apiVersion: apps/v1
 kind: Deployment
@@ -57,3 +58,8 @@ spec:
         value: arm64
         effect: NoSchedule
       {{end}}
+      {{if $TOLERATION}}
+      - key: {{$TOLERATION}}
+        operator: Exists
+        effect: NoSchedule
+      {{end}}