WorkflowReconciled fails with DNS resolution error when operator notifies runner pod

[ambient-code]

Category: bug / infrastructure

Summary

The operator fails to deliver workflow files to the runner pod due to a DNS resolution failure when looking up the runner pod's internal cluster hostname. The runner pod starts successfully (RunnerStarted: True) but the workflow is never loaded because the operator cannot reach it.

Error

Failed to notify runner: Post "http://session-dep-bump-agentic-sandbox-repo-1777982576.ols-shared.svc.cluster.local:8001/workflow": dial tcp: lookup session-dep-bump-agentic-sandbox-repo-1777982576.ols-shared.svc.cluster.local on 172.30.0.10:53: no such host

Condition State

WorkflowReconciled:
  Status: False
  Reason: UpdateFailed
  Message: Failed to notify runner: <DNS error above>

RunnerStarted: True

Steps to Reproduce

1. Create a session with a workflow configuration
2. Session runner pod starts successfully
3. Operator attempts to POST workflow files to the runner at its internal cluster hostname
4. Result: DNS lookup for the runner pod hostname fails with no such host
5. Workflow files are never loaded into the workspace; session is broken

Expected Behavior

Operator should successfully resolve the runner pod hostname and deliver workflow files after RunnerStarted is True.

Actual Behavior

• Runner pod is running and healthy
• Operator cannot resolve the internal cluster DNS name for the runner pod
• WorkflowReconciled condition stays False with UpdateFailed
• Session is left in a broken state — workflow never loads

Notes

• DNS server is 172.30.0.10:53 (cluster DNS)
• The hostname pattern is session-{name}.{namespace}.svc.cluster.local — possible mismatch between the service name the operator constructs and the actual service created for the pod
• This is a platform/infrastructure issue, not a workflow configuration issue
• May be a timing issue (operator tries to notify before the service/DNS entry is ready) or a naming mismatch

---

Filed via Amber Interview