Fix role deletion & change to use p99.9 for Pod Identity latency

Author: xdu31

tests/tekton-resources/pipelines/eks/awscli-cl2-load-with-addons-slos.yaml

@@ -35,6 +35,7 @@ spec:
   - name: desired-nodes
     type: string
   - name: pods-per-node
+    default: "50"
     type: string
   - name: nodes-per-namespace
     type: string
@@ -77,13 +78,13 @@ spec:
   - name: pia-test-pod-spec-url
     default: "https://raw.githubusercontent.com/awslabs/kubernetes-iteration-toolkit/main/tests/assets/eks-pod-identity/pod-default.yaml"
   - name: cl2-eks-pod-identity-pods
-    default: "5000"
+    default: "10000"
   - name: cl2-default-qps
-    default: "100"
+    default: "300"
   - name: cl2-default-burst
-    default: "200"
+    default: "300"
   - name: cl2-uniform-qps
-    default: "100"
+    default: "300"
   - name: cl2-metric-dimension-name
     description: "default metric dimension name"
     default: "ClusterName"
@@ -97,7 +98,7 @@ spec:
     description: "default metric period"
     default: "300"
   - name: timeout-pia-pod-creation
-    default: "80s"
+    default: "90s"
   - name: timeout-pia-pod-startup
     default: "60s"
   - name: launch-template-ami

tests/tekton-resources/tasks/generators/clusterloader/load-pod-identity.yaml

@@ -228,13 +228,13 @@ spec:
         --start-time "$START_TIME" \
         --end-time "$END_TIME" \
         --period "$PERIOD" \
-        --extended-statistics p50 p99 p99.95 \
+        --extended-statistics p99 p99.9 p99.95  \
         --output json)
 
-      # extract p50 p99 p99.95 of credential fetching
+      # extract p99 p99.9 p99.95 of credential fetching
       latest=$(echo "$response" | jq -r '.Datapoints | sort_by(.Timestamp) | last')
-      p50=$(echo "$latest" | jq -r '."ExtendedStatistics"."p50" // "N/A"')
       p99=$(echo "$latest" | jq -r '."ExtendedStatistics"."p99" // "N/A"')
+      p999=$(echo "$latest" | jq -r '."ExtendedStatistics"."p99.9" // "N/A"')
       p9995=$(echo "$latest" | jq -r '."ExtendedStatistics"."p99.95" // "N/A"')
 
       response=$(aws cloudwatch get-metric-statistics \
@@ -258,8 +258,8 @@ spec:
         "end_time": "$END_TIME",
         "total_samples": $total_samples,
         "rate": $rate,
-        "p50": $p50,
         "p99": $p99,
+        "p99.9": $p999,
         "p99.95": $p9995
       }
       EOF
@@ -268,10 +268,10 @@ spec:
       ls -larth
       aws s3 cp . s3://$S3_RESULT_PATH/  --recursive
 
-      # if p99.95 is equal to or more than 1 second, exit with failure
-      int_p9995=$(echo "$p9995" | awk '{printf "%d", $1}')
-      echo "p99.95 is $p9995"
-      if [ "$int_p9995" -lt 1 ]; then
+      # if p99.9 is equal to or more than 1 second, exit with failure
+      int_p999=$(echo "$p999" | awk '{printf "%d", $1}')
+      echo "p99.9 is $p999"
+      if [ "$int_p999" -lt 1 ]; then
         echo "1" | tee $(results.datapoint.path)
       else
         echo "0" | tee $(results.datapoint.path)

tests/tekton-resources/tasks/teardown/awscli-eks.yaml

@@ -53,13 +53,13 @@ spec:
         if [ "$PIA_ROLE_EXISTS" == "$PIA_ROLE_NAME" ]; then
           # Detach all attached managed policies
           aws iam list-attached-role-policies --role-name "$PIA_ROLE_NAME" \
-            --query 'AttachedPolicies[*].PolicyArn' --output text | while read -r policy_arn; do
+            --query 'AttachedPolicies[*].PolicyArn' --output json | jq -r '.[]' | while read -r policy_arn; do
             echo "Detaching managed policy: $policy_arn"
             aws iam detach-role-policy --role-name "$PIA_ROLE_NAME" --policy-arn "$policy_arn"
           done
           # Delete all inline policies
           aws iam list-role-policies --role-name "$PIA_ROLE_NAME" \
-            --query 'PolicyNames' --output text | while read -r policy_name; do
+            --query 'PolicyNames' --output json | jq -r '.[]' | while read -r policy_name; do
             echo "Deleting inline policy: $policy_name"
             aws iam delete-role-policy --role-name "$PIA_ROLE_NAME" --policy-name "$policy_name"
           done