Supported combination: platform 2026.10.1rc17 · tooling 2026.10.1rc17.
Run and recover
Troubleshooting
Locate failures across connection, deployment, preparation, and execution.
Start with the failing stage. A successful local validation proves document shape, not server access, Action availability, or external service permissions.
| Symptom | Check |
|---|---|
| Certificate verification fails | Server hostname, system trust, and tls.ca_file; see configuration. |
| Login fails or a request is unauthorized | Server URL, account status, and session/token validity. Log in again when the session expires. |
| Permission denied | Your Environment-scoped role; starting, deploying, approving, and administering require different permissions. |
| Action cannot be resolved during deployment | Registry host, repository prefix, credentials, exact Action version, and target platform. |
| Deployment revision conflict | Read the latest revision and review the intervening change before redeploying. |
| Preparation blocks | Missing context/secrets, incompatible input types, invalid references, or missing outputs from a skipped step. |
| Run waits for approval | Review the prepared step and obtain eligible approvals under all applicable policies. |
| Run remains queued | Read its diagnostic; check Environment occupancy, runner admission/connectivity, capacity, and image startup. |
| Action blocks after starting | Inspect attempt outcome, exit code, errors, external permissions, and non-private logs. |
| Logs or outputs are absent | Check whether the step is private, execution started, and logs reached their storage bounds. |
| Run becomes uncertain | Restore the owning runner and let it reconcile; see recovery. |
Collect useful evidence
Terminal
pyplines --json run info 42
pyplines --json run attempts 42
pyplines --json run history 42
Include the platform/tooling versions, Environment, Pypline revision, Run ID, step name, and invocation ID when escalating an issue. Inspect outputs before sharing them; ordinary parameters and logs can still contain operational details.
For deployment failures, preserve the Definition and the full error. For runner issues, an administrator should also inspect pyplines runner list and the owning runner's service logs. Avoid modifying server state or discarding runner state to make a diagnostic disappear.

