Troubleshooting

The following section provides a quick reference for common issues, their causes, and actions.

• Pods remain in the Pending state.

  • Likely cause: The NodePool is not ready or does not have sufficient capacity.
  • Action: Check the NodePool status and verify that sufficient capacity is available.

• HPA displays unknown metrics.

  • Likely cause: The Metrics Server is missing or unhealthy.
  • Action: Install the Metrics Server or restore its health.

• Gateway API returns a 401 Unauthorized response.

  • Likely cause: JWT authentication fails because the token is missing, malformed, or expired.
  • Action:
    • Verify that the Authorization: Bearer <JWT_TOKEN> header is present and valid.
    • If required, Verify application functionality without authentication.

• Uninstall stops responding.

  • Likely cause: A Karpenter finalizer prevents resource deletion.
  • Action:Manually remove the remaining resources

• NodePool or NodeClass resources remain in the cluster.

  • Likely cause: Insufficient permissions prevent resource deletion.
  • Action:Manually remove the remaining resources

Verify application functionality without authentication

kubectl -n data-discovery run curl --image=curlimages/curl -it --rm --restart=Never -- \
curl -v -X POST classification-service:8050/pty/data-discovery/v2/classify/text \
-H 'Content-Type: text/plain' \
--data 'Detect Jane Roe phone 203-555-1111'

Manually remove the remaining resources

Follow the steps to manually remove resources.

1. Remove finalizer from the Data Discovery EC2NodeClass.

   kubectl patch ec2nodeclass data-discovery-nodeclass \
   --type merge \
   -p '{"metadata":{"finalizers":[]}}'
   kubectl delete ec2nodeclass data-discovery-nodeclass
   

2. Delete NodePools.

   kubectl delete nodepool data-discovery-classification
   kubectl delete nodepool data-discovery-context  
   kubectl delete nodepool data-discovery-pattern
   

3. Delete EC2NodeClass (AWS provider)

   kubectl delete ec2nodeclass data-discovery-nodeclass