Runtime diagnostics and health
Entry and workflow
Diagnostics; workspace node details. All three roles can read diagnostic summaries. Diagnostic switches are deployment settings; viewing this page does not grant server configuration access. See the role matrix for all actions.
Configuration and actions
Runtime diagnostics and node status are separate pages. Diagnostic switches are deployment configuration; viewing these pages does not grant server-configuration access.
Runtime diagnostic information
| Display | Meaning | Shown when |
|---|---|---|
| Expired leases | A work item remains CLAIMED and its lease deadline is at/before database observation time. | Always |
| Queue delays | READY/RETRY_WAIT items exceed the configured wait threshold; a match does not establish an eligible Worker exists. | Always |
| Root deadlines | A root exceeds its deadline while nonterminal; children and terminal runs are excluded. | Always |
| Unsettled commands | A command expired while still PENDING/PROCESSING. | Always |
| Observed condition | Correlate state, target and time; this alone does not prove Worker failure. | Always |
| Node and component state | On node/runtime status pages inspect roles, online state and sample freshness; unknown is not zero. | Node/runtime status pages |
| Queue and capacity | Interpret available waiting, in-flight, thread and connection metrics; missing data do not mean no load. | Node/runtime status pages |
Actions
| Action | Roles | Available when | Result |
|---|---|---|---|
| Refresh checks | ADMIN / OPERATOR / VIEWER | Summary access and valid check conditions | Read current diagnostic facts without automatically stopping or retrying tasks. |
| View evidence | ADMIN / OPERATOR / VIEWER | The check contains a matching record | Inspect timestamps and related objects; open run details for logs, which require ADMIN. |
| View root run | ADMIN / OPERATOR / VIEWER | Evidence includes a root-instance reference | Verify live state and changes; a diagnostic snapshot does not replace the run outcome. |
Use and verify
Establish a baseline with docs-hello, then inspect the intentional failure. Diagnostics support investigation but do not replace target-system verification. Follow Troubleshooting after database recovery/node changes.