docs(docker): dashboard IS supervised — update note that contradicted the PR

PR #30136 review caught that website/docs/user-guide/docker.md still
said "The dashboard side-process is **not supervised** — if it
crashes, it stays down until the container restarts." That was true
under tini but is the opposite of the s6 behavior this PR ships and
`test_dashboard_restarts_after_crash` proves.

Replace with a description of what users actually see now: automatic
restart by s6-overlay, new PID after a short backoff, logs via
`docker logs`. The standalone-container caveat carries forward
unchanged.
This commit is contained in:
Ben 2026-05-23 15:08:48 +10:00
parent efd3569739
commit 04d1894f36

View file

@ -85,7 +85,15 @@ The entrypoint starts `hermes dashboard` in the background (running as the non-r
The default `HERMES_DASHBOARD_HOST=0.0.0.0` is required for the host to reach the dashboard through the published port; the entrypoint automatically passes `--insecure` to `hermes dashboard` in that case. Override to `127.0.0.1` if you want to restrict the dashboard to in-container access only (e.g. behind a reverse proxy in a sidecar).
:::note
The dashboard side-process is **not supervised** — if it crashes, it stays down until the container restarts. Running it as a separate container is not supported: the dashboard's gateway-liveness detection requires a shared PID namespace with the gateway process.
The dashboard runs as a supervised s6 service inside the container. If
the dashboard process crashes, s6-overlay restarts it automatically
after a short backoff — you'll see a new PID without needing to
restart the container. Logs and crash output are visible via
`docker logs <container>` (s6 forwards service stdout/stderr there).
Running the dashboard as a separate container is not supported: its
gateway-liveness detection requires a shared PID namespace with the
gateway process.
:::
## Running interactively (CLI chat)