Deploy on Railway Keeps Failing

[azizpunjani]

Hello! I'm trying to deploy to Railway to test mcp-automem out. The deploy keeps failing.

[jack-arturo]

@azizpunjani Thanks for the screenshot — this makes the failure mode clearer.

This does not look like a build failure. The image appears to build successfully, and Railway is failing later during the runtime healthcheck phase:

• build completes
• Railway starts the container
• Railway probes GET /health
• the service never becomes healthy
• deploy fails with 1/1 replicas never became healthy

So the likely issue is one of:

1. the service is not listening on Railway’s assigned PORT
2. the service is not binding to 0.0.0.0
3. the service does not expose GET /health
4. the process is crashing on startup or blocking before it starts listening

The next thing that would help most is the Deploy Logs tab rather than the Build Logs tab.

If you can, please share:

• the Deploy Logs
• the Dockerfile
• the server.js you’re deploying

The most common fix on Railway is making sure startup looks roughly like this:

const port = process.env.PORT || 3000;
app.listen(port, "0.0.0.0");
app.get("/health", (_req, res) => res.status(200).send("ok"));

If the service waits for upstream dependencies before binding the port, that can also cause Railway healthchecks to fail even when the code itself is otherwise fine.

[jack-arturo]

@azizpunjani Quick update — there's a strong chance this is now fixed.

Last week I hit the same 1/1 replicas never became healthy failure shipping release 0.15.1 on Railway. Root cause: the mcp-automem SSE bridge's /health endpoint was returning HTTP 503 whenever upstream FalkorDB↔Qdrant had any sync drift (which is normal under write load). Railway treats 503 as unavailable and kills the deploy.

Fixed in automem#151 (merged Apr 24): /health now returns 200 whenever the bridge process is up, and strict upstream gating moved to a new /ready endpoint. The fix is in ghcr.io/verygoodplugins/mcp-automem:stable as of Apr 25.

If you used the Railway template button:

1. Open the failing service in Railway
2. Deployments → Deploy latest to pull the new :stable image
3. Healthcheck should pass

If you wrote your own Dockerfile, rebase against main on verygoodplugins/automem (or pin to :v0.15.3+).

Could you give it another try and let me know if it deploys cleanly? If it still fails, please share the Deploy Logs tab (not Build Logs) — that'll tell us if it's a different issue.