Merge pull request #3530 from MicrosoftDocs/repo_sync_working_branch

Confirm merge from repo_sync_working_branch to main to sync with https://github.com/MicrosoftDocs/azure-ai-docs (branch main)

Author: Taojunshen

articles/machine-learning/how-to-troubleshoot-kubernetes-extension.md

@@ -78,6 +78,54 @@ When you request support, we recommend that you run the following command and se
 ```bash
 kubectl logs healthcheck -n azureml
 ```
+## Extension-operator pod in azure-arc/kube-system namespace is crashing due to OOMKill 
+This issue happens if the extension's helm chart size is large and there are multiple Helm releases on the cluster. Here is a sample script to help clean up the helm history on the cluster: 
+```
+#!/bin/bash
+
+# Set release name and namespace
+RELEASE_NAME=$1
+NAMESPACE=$2
+
+# Validate input
+if [[ -z "$RELEASE_NAME" || -z "$NAMESPACE" ]]; then
+    echo "Usage: $0 <release-name> <namespace>"
+    exit 1
+fi
+
+echo "Fetching Helm history for release: $RELEASE_NAME in namespace: $NAMESPACE"
+
+# Get stuck revisions (PENDING_ROLLBACK or PENDING_UPGRADE) using grep + awk for accurate parsing
+STUCK_REVISIONS=$(helm history "$RELEASE_NAME" -n "$NAMESPACE" | grep 'pending-' | awk '{print $1}')
+
+if [[ -z "$STUCK_REVISIONS" ]]; then
+    echo "No stuck Helm revisions found. Nothing to delete."
+    exit 0
+fi
+
+echo "Found stuck Helm revisions: $STUCK_REVISIONS"
+
+# Loop through each stuck revision and delete the corresponding secret
+for REVISION in $STUCK_REVISIONS; do
+    SECRET_NAME="sh.helm.release.v1.${RELEASE_NAME}.v${REVISION}"
+   
+    echo "Deleting Helm history secret: $SECRET_NAME"
+   
+    kubectl delete secret -n "$NAMESPACE" "$SECRET_NAME" --ignore-not-found
+done
+
+echo "Cleanup complete. Verify with 'helm history $RELEASE_NAME -n $NAMESPACE'"
+
+exit 0
+
+```
+
+How to run the script: 
+```
+chmod +x delete_stuck_helm_secrets.sh
+
+./delete_stuck_helm_secrets.sh my-release my-namespace
+```
 
 ### Error Code of HealthCheck 
 This table shows how to troubleshoot the error codes returned by the HealthCheck report.