Operator Operations

Copy Markdown View Source

This guide is for day-2 usage: verify, observe, and troubleshoot.

Verify resources

kubectl get flamepools -A
kubectl get flamerunners -A
kubectl get pods -A
kubectl get events -A --sort-by=.lastTimestamp

Inspect runner status

kubectl get flamerunner -n <ns> <name> -o yaml

Focus on:

  • status.phase
  • status.reason
  • status.message
  • status.conditions

Inspect pool scheduling feedback

kubectl get flamepool -n <ns> <name> -o yaml

Focus on:

  • status.resolvedScheduling
  • status.schedulingFeedback
  • status.conditions

Common checks

  1. CRDs established:
kubectl get crd flamepools.flame.org flamerunners.flame.org
  1. Operator healthy:
kubectl get deploy -n flame
kubectl get pods -n flame
kubectl logs -n flame deploy/flame-k8s-controller
  1. Webhook registered:
kubectl get mutatingwebhookconfigurations | grep flame

Upgrade strategy

Use release tags and install script with explicit version:

bash /tmp/install-operator.sh --tag v0.1.0 --namespace flame

Pinning versions avoids accidental behavior drift.

Safe cleanup

Delete application workloads first, then pools/runners if required. Finalizers protect against inconsistent deletion order.