fix: cleanup code and configurable throughput-pods

Author: dheeraj-coding

tests/tekton-resources/pipelines/eks/awscli-eks-cl2-steady-load.yaml

@@ -15,6 +15,7 @@ spec:
     description: "scheduler throughput threshold"
   - name: cl2-scheduler-steady-qps
     description: "expected PCP tier scheduler qps"
+  - name: cl2-scheduler-throughput-pods
   - name: cl2-scheduler-create-qps
   - name: results-bucket
   - name: slack-hook
@@ -181,12 +182,16 @@ spec:
       value: $(params.cl2-scheduler-steady-qps)
     - name: cl2-scheduler-create-qps
       value: $(params.cl2-scheduler-create-qps)
+    - name: cl2-scheduler-throughput-pods
+      value: $(params.cl2-scheduler-throughput-pods)
     - name: nodes
       value: $(params.desired-nodes)
     - name: results-bucket
       value: $(params.results-bucket)
     - name: cluster-name
       value: $(params.cluster-name)
+    - name: control-plane-tier
+      value: $(params.control-plane-tier)
     runAfter:
     - create-mng-nodes
     taskRef:

tests/tekton-resources/tasks/generators/clusterloader/sustained-scheduler-throughput.yaml

@@ -13,8 +13,6 @@ spec:
   - name: cl2-branch
     description: "The branch of clusterloader2 you want to use"
     default: "steady-load"
-  - name: cl2-scheduler-throughput-pods
-    default: 500000
   - name: cl2-scheduler-throughput-threshold
     description: "scheduler throughput threshold"
     default: "150"
@@ -33,6 +31,8 @@ spec:
     description: The region where the cluster is in.
   - name: cluster-name
     description: The name of the EKS cluster you want to spin.
+  - name: control-plane-tier
+    default: "standard"
   results:
     - name: datapoint
       description: Stores the CL2 result that can be consumed by other tasks (e.g. cloudwatch) 
@@ -83,17 +83,57 @@ spec:
     script: |
       #!/bin/bash
 
-      TOTAL_CYCLES=1
+      TIER=$(params.control-plane-tier)
+      ITERATIONS=0
+
+      case $TIER in
+        "tier-xl")
+          ITERATIONS=4
+          ;;
+        "tier-2xl")
+          ITERATIONS=8
+          ;;
+        "tier-4xl")
+          ITERATIONS=16
+          ;;
+        *)
+          echo "Unknown tier: $TIER. Defaulting to 1 iteration."
+          ITERATIONS=1
+          ;;
+      esac
+
       cat $(workspaces.source.path)/perf-tests/clusterloader2/testing/scheduler-throughput/config.yaml
       cd $(workspaces.source.path)/perf-tests/clusterloader2/
 
-      ENABLE_EXEC_SERVICE=false ./clusterloader --kubeconfig=$KUBECONFIG --testconfig=$(workspaces.source.path)/perf-tests/clusterloader2/testing/sustained-scheduler-throughput/config.yaml --testoverrides=$(workspaces.source.path)/overrides.yaml --nodes=$(params.nodes) --provider=eks --report-dir=$(workspaces.results.path) --alsologtostderr --v=2
+      FAIL_COUNT=0
+      for ((i=1; i<=ITERATIONS; i++)); do
+        echo "--- Starting iteration $i of $ITERATIONS ---"
+
+        ENABLE_EXEC_SERVICE=false ./clusterloader --kubeconfig=$KUBECONFIG --testconfig=$(workspaces.source.path)/perf-tests/clusterloader2/testing/sustained-scheduler-throughput/config.yaml --testoverrides=$(workspaces.source.path)/overrides.yaml --nodes=$(params.nodes) --provider=eks --report-dir=$(workspaces.results.path) --alsologtostderr --v=2
+
+        if [ $? -ne 0 ]; then
+          echo "Iteration $i failed."
+          ((FAIL_COUNT++))
+        else
+          echo "Iteration $i passed."
+        fi
+      done
+
+      FAILURE_RATE=$(( (FAIL_COUNT * 100) / ITERATIONS ))
+
+      echo "------------------------------------------"
+      echo "Total Iterations: $ITERATIONS"
+      echo "Total Failures:   $FAIL_COUNT"
+      echo "Failure Rate:     $FAILURE_RATE%"
+      echo "------------------------------------------"
 
-      exit_code=$?
-      if [ $exit_code -eq 0 ]; then
-        echo "1" | tee $(results.datapoint.path)
+      exit_code=0
+      if [ "$FAILURE_RATE" -ge 25 ]; then
+        echo "Error: Failure rate ($FAILURE_RATE%) exceeded the 25% threshold."
+        exit_code=1
       else
-        echo "0" | tee $(results.datapoint.path)
+        echo "Success: Failure rate ($FAILURE_RATE%) is within acceptable limits."
+        exit_code=0
       fi
       ls -a $(workspaces.results.path)
       cat $(workspaces.results.path)/SchedulingThroughput*

tests/tekton-resources/tasks/setup/eks/awscli-cp-with-pcp.yaml

@@ -12,7 +12,7 @@ spec:
   - name: cluster-name
     description: The name of the EKS cluster you want to spin.
   - name: kubernetes-version
-    default: "1.23"
+    default: "1.33"
     description: The EKS version to install.
   - name: region
     default: "us-west-2"
@@ -27,9 +27,6 @@ spec:
   - name: aws-ebs-csi-driver-version
     default: release-1.53
     description: The release version for aws ebs csi driver.
-  - name: aws-pod-identity-agent-version
-    default: v1.3.9-eksbuild.1
-    description: The release version for aws pod identity agent.
   - name: control-plane-tier
     default: "standard"
     description: "Provisioned Control Plane tier, default to standard"
@@ -116,10 +113,3 @@ spec:
         }
       }' -n kube-system
       kubectl scale --replicas 1000 deploy coredns -n kube-system
-      # Install EKS Pod Identity Agent
-      ENDPOINT_FLAG=""
-      if [ -n "$(params.endpoint)" ]; then
-        ENDPOINT_FLAG="--endpoint $(params.endpoint)"
-      fi
-      aws eks $ENDPOINT_FLAG create-addon --cluster-name $(params.cluster-name) --addon-name eks-pod-identity-agent --addon-version $(params.aws-pod-identity-agent-version)
-      aws eks $ENDPOINT_FLAG --region $(params.region) wait cluster-active --name $(params.cluster-name)