Supabase template: `rest` and `functions` have no healthcheck (service stays grey in EasyPanel/Swarm)

[Carlos-Vera]

Summary

In supabase/code/docker-compose.yml (branch main), two services run without any healthcheck:

• rest → postgrest/postgrest:v12.2.12
• functions → supabase/edge-runtime:v1.67.4

kong and meta also omit a Compose-level healthcheck, but their images ship a HEALTHCHECK, so Docker still reports them as healthy. rest and functions have none at either level.

Why it matters

EasyPanel aggregates per-container health into a single status for the Compose service. A container with no healthcheck is reported by Docker as plain Up (no health state), so EasyPanel keeps the whole service grey and never turns it green, even when everything is fully working. Operationally you cannot tell "no check defined" from "actually down".

Verified on a running deployment (EasyPanel on Docker Swarm):

$ docker ps --format 'table {{.Names}}\t{{.Status}}' | grep supabase
...-rest-1        Up 15 hours
...-functions-1   Up 15 hours
...(other 11)     Up 15 hours (healthy)

Proposed fix

functions (supabase/edge-runtime) — the image ships bash and the runtime listens on :9000 (the port Kong proxies /functions/v1 to). A dependency-free TCP check works:

    healthcheck:
      test: [ "CMD", "bash", "-c", "exec 3<>/dev/tcp/127.0.0.1/9000" ]
      interval: 10s
      timeout: 5s
      retries: 3
      start_period: 20s

rest (postgrest/postgrest) — this image is distroless (no sh/bash/curl/wget), so a standard CMD/CMD-SHELL check cannot run inside it. PostgREST exposes a built-in admin server with /live and /ready when PGRST_ADMIN_SERVER_PORT is set, but there is no in-image tool to query it. Options: mount a tiny static healthcheck helper, document the limitation, or rely on the upstream PostgREST image gaining a HEALTHCHECK.

Happy to send a PR with the functions fix.

Notes

The same gap exists upstream in supabase/supabase (docker/docker-compose.yml); also reporting it there.

[Carlos-Vera]

Filed upstream in Supabase as well: supabase/supabase#46642

[Carlos-Vera]

Quick update — both healthchecks are now merged upstream in supabase/supabase master:

• rest (PostgREST): fix(self-hosted): add healthcheck to rest (PostgREST) service supabase/supabase#46658
• functions (edge-runtime): fix(docker): add healthcheck to edge functions container supabase/supabase#46655

Since this template mirrors supabase/docker via update.js, they'll come down on the next sync. One heads-up: the template is a fair bit behind master right now (postgrest v12.2.12 here vs v14.12 upstream), so a regen pulls in a year's worth of Supabase changes, not just the healthchecks — and the rest one specifically needs PostgREST v14 (postgrest --ready), so it only works once the image bump lands with it. I checked that the removePorts/removeContainerNames steps don't touch the new healthcheck/environment blocks, so the fix survives the transform cleanly.

Happy to send the sync PR if that's useful, or leave the version bump to you — your call, since it's a bigger update with a blast radius on existing deployments.

[Carlos-Vera]

Sent a PR for this: #79, against 18-05-2026. Just the two healthcheck blocks for rest and functions, no image bumps.

The rest one is the same fix already merged upstream in supabase/supabase#46658 — PostgREST is distroless, so the check uses its own postgrest --ready binary against the admin server on localhost, with the port left unpublished.

Running it on EasyPanel for ~22 hours, all 13 containers healthy.

Separately, the grey service indicator turned out to be a panel-side bug, confirmed by @Ahson-Shaikh in Discord — not something this template can fix.