SkyfloDocs
OperationsTroubleshooting

Troubleshooting

Common issues and their solutions.

Pods Not Starting

Check pod status and events:

bash
kubectl get pods -n skyflo
kubectl describe pod <pod-name> -n skyflo

Common causes:

  • ImagePullBackOff: Verify imagePullSecrets if using a private registry.
  • Pending: Check node resources. Skyflo requires approximately 800m CPU and 1.8Gi memory total.
  • CrashLoopBackOff: Check logs with kubectl logs <pod-name> -n skyflo.

Engine Cannot Connect to PostgreSQL

Verify the database pod is running:

bash
kubectl get pods -n skyflo -l app=skyflo-postgresql

If using an external database, confirm the connection URL in engine.secrets or postgresql.external.url.

Engine Cannot Connect to Redis

bash
kubectl get pods -n skyflo -l app=skyflo-redis

If using external Redis, verify redis.external.url in your values.

Command Center (ui) Not Loading

  1. Confirm the ui pod is running: kubectl get pods -n skyflo -l app=skyflo-ui
  2. Check port-forward is active: kubectl port-forward svc/skyflo-ui -n skyflo 3000:80
  3. Verify Engine CORS allows the Command Center origin.

Chat Not Streaming

If the chat interface connects but does not stream responses:

  • Verify NEXT_PUBLIC_API_URL points to the Engine.
  • Check Engine logs for errors: kubectl logs -n skyflo -l app=skyflo-engine --tail=100
  • Confirm the LLM provider API key is set and valid.
  • Check rate limiting: rateLimitPerMinute defaults to 100 requests.

Approval Stream Returns 404

Ensure the call_id in the approval request matches a pending tool call. Approval streams are ephemeral and expire after the tool call is resolved.

MCP Tools Not Discovered

The Engine discovers MCP tools on startup. If tools are missing:

bash
kubectl logs -n skyflo -l app=skyflo-mcp --tail=50

Verify the MCP server is running and the mcpServerUrl is correctly resolved.

Jenkins Integration Failures

  • Verify the Kubernetes Secret exists and contains username and api-token keys.
  • Confirm the Jenkins API URL is reachable from inside the cluster.
  • Check for CSRF issues in Jenkins logs.

See Integrations for credential setup details.

Getting Help