Update finished, but Doctor says the agent database is still in use

Troubleshooting
, , , ,
1

Adapted community discussion: Paraphrased and condensed from an anonymized public OpenClaw Discord support conversation from September 2026. Technical guidance was reviewed for OpenClaw 2026.9.3.

I ran the supported updater on Windows after a clean dry run. The package moved to the stable release, but the post-update Doctor step failed because several agent databases still need a stopped-writer migration. The Gateway is now unreachable, so the command reports the overall update as an error even though it also says the new version was installed.

Should I run the update again, reinstall, or repair the database somehow?

2

The package update completed. Do not rerun it or delete any SQLite files. The remaining work is the stopped-writer migration.

Close the Windows Hub and any agent windows, then use a fresh PowerShell window:

openclaw gateway stop
openclaw doctor --fix

Only after Doctor completes without the database-readiness error, finish the interrupted update and restart the service:

openclaw update repair
openclaw gateway restart
openclaw --version
openclaw gateway status --deep

If Doctor still reports an active writer, another local OpenClaw process is holding the database. Preserve the database and all of its sidecar files while locating that process.

3

The Gateway Scheduled Task says it stopped, but openclaw gateway status --deep still shows the same pending agent-database migration. The service is not running and the connectivity probe is refused.

4

That confirms the Scheduled Task is stopped, while another local OpenClaw process still owns the database. The safest Windows path is a reboot. After signing in, do not open the Hub or any agent. Run openclaw gateway stop once more, then openclaw doctor --fix.

If you prefer to avoid a reboot, open Resource Monitor, search Associated Handles for openclaw-agent.sqlite, and close only the identified OpenClaw process. Avoid killing every Node process on the machine.

5

I am still getting the same migration loop after stopping the Gateway.

6

Doctor appears to have restarted the Scheduled Task during its previous attempt, which reintroduced the writer. Stop the Gateway, reboot, and run Doctor before opening any OpenClaw UI. Do not run openclaw update repair until Doctor finishes without the schema or database-owner error.

This thread did not capture a confirmed recovery, so treat the sequence as a safe diagnostic path rather than a verified solution. Current reference: Doctor.