Possible to increase stop/restart timeout?

[HerbCSO]

I am running an Omada Controller container that is currently having issues with shutting down properly (see mbentley/docker-omada-controller#807), so it takes ~30s to come down and it exits with a code 137 (killed). That 30s happens to be long enough that Dockhand thinks it has failed to stop the container, so when it goes to do an update, it tells it to stop, that takes longer than 30s, Dockhand aborts the update process, and I'm left with a non-running container.

This is what that looks like in the logs:

[dockhand-dockhand-1] 2026-09-14T08:09:10.596401882-04:00 2026-09-14T12:09:10.596Z INFO  [EnvUpdateCheck]   Recreating container...
[dockhand-dockhand-1] 2026-09-14T08:09:10.618189584-04:00 2026-09-14T12:09:10.618Z INFO  [EnvUpdateCheck]   Recreating container: omada-controller (image: mbentley/omada-controller:6)
[dockhand-dockhand-1] 2026-09-14T08:09:10.618202096-04:00 2026-09-14T12:09:10.618Z INFO  [EnvUpdateCheck]   Stopping container...
[dockhand-dockhand-1] 2026-09-14T08:09:40.619414942-04:00 2026-09-14T12:09:40.619Z ERROR [Docker] hawser-standard <hostname_redacted>: POST /containers/81a0c0558ce4e4175be56a045de1dae2887c00eaa9442142f1762ec7335d4715/stop failed after 30001ms: The operation was aborted due to timeout
[dockhand-dockhand-1] 2026-09-14T08:09:40.619531474-04:00 2026-09-14T12:09:40.619Z INFO  [EnvUpdateCheck]   Failed to recreate container: Connection timeout
[dockhand-dockhand-1] 2026-09-14T08:09:40.619759447-04:00 2026-09-14T12:09:40.619Z INFO  [EnvUpdateCheck]   FAILED: Connection timeout

Aside: as you can tell, this is happening with a Hawser agent on the host - not sure if that makes a difference, I don't have a similar container running on the same host as Dockhand itself, so I don't know if that would behave the same way.

Now, I know that the real solution is a fix in the Omada controller container so that it doesn't misbehave like this anymore, but that actually requires a fix from TPLink in the underlying Omada Controller software, so that could take a while (there has been discussion in the TPLink forums about this already).

In the meantime I'd like to know if there is a way to increase this timeout. It doesn't need to be much, since the container does come down in just over 30s, and there actually occasions where this does succeed when it happens to sneak in JUST under the timeout. I'd love to be able to do this as a temporary workaround so I don't have to keep going in and restarting this container when Uptime Kuma tells me it's down again... ;]

[sueun-dev]

COMPOSE_TIMEOUT doesn't extend this individual stop request in the current plain-HTTP implementation. At commit 6963f7d, /containers/<id>/stop gets a fixed 30-second request deadline, whereas COMPOSE_TIMEOUT applies to /_hawser/compose. The manual also describes that variable as a Compose-operation timeout.

With COMPOSE_TIMEOUT=60, I exercised the extracted dockerFetch function on Node 22.23.2 against local HTTP responses delayed by 31 seconds: stop aborted at 30 seconds, while Compose returned 204. An immediate stop response succeeded too. This was an isolated function test, not a Hawser/Omada integration test.

Which Dockhand and Hawser versions are you running, and is the agent connection HTTP or HTTPS? That will establish whether this is the same request path before looking for a version-specific workaround. HTTPS uses a separate implementation.

[HerbCSO]

Dockhand: v1.0.48
Hawser: 0.2.47
Connection is HTTP right now (although I might want to switch to HTTPS but I'd have to use a self-signed cert I think and I haven't worked through how to set that up yet)

[HerbCSO]

Is there anything else you need on this?

[sueun-dev]

Sorry for the slow follow-up! No additional details needed from you at this point; the versions and HTTP connection detail were enough to check the relevant code path.

In [v1.0.48](https://github.com/Finsys/dockhand/blob/v1.0.48/src/lib/server/docker.ts#L968-L974), that stop request has a hard-coded 30-second timeout. I also checked [v1.0.53](https://github.com/Finsys/dockhand/blob/v1.0.53/src/lib/server/docker.ts#L951-L957), and the same limit is still there. COMPOSE_TIMEOUT doesn't apply to this request.

I don't have a tested workaround for your setup. Exposing a configurable timeout for this request would require a change in Dockhand.

[HerbCSO]

No worries, I'll hang tight, thank you.