503 means usually means your app didn't respond to the request from Railway's proxy, have you checked your deployment logs?

Are you listening on 0.0.0.0 and $PORT?

Yes! Listening on 0.0.0.0 and $PORT. No errors appear in the logs, and it's handling 98% of the requests, just some seem to fail

Are you getting these errors at the same time you are deploying?

Nope, after deploying

Shortly after deploying?

Let's say > 15 minutes after deploying

Did you change a variable >15 minutes after deploying?

No, nothing

All signs are currently pointing to an issue with the app, tell me more about what the failing request does internally including any middleware.

The only thing that it does is return data (where data is a variable). No middleware, no validation. That's why I think that it is almost impossible for the app to not be able to respond

After this >15 minute mark, if you hit the app at 1 req/s for 5 minutes, how many requests would you say fail?

I'd say 5 out of 300. It is very occasionally, but still want to make things sure :)

What is the start command in use?

No custom start command is set in the config, so I assume it checked my package.json and did yarn start (as start is defined in the package.json)

Right and what is yarn start set to.

nodemon start src/index.js

Regardless if nodemon is causing issues here, it's never a good idea to use nodemon in production, remove the use of nodemon and let me know if the issue continue.

I know, I am testing out railway, so I just kept everything as-is. For production I wouldn't use it either. The issue happens after the app has been started, so that shouldn't be related to Nodemon, right?

Even locally you shouldn't use nodemon in the start script, nodemon should only be used in dev scripts and alike.

At this time I wouldn't know if nodemon is causing this issue, but it's definitely incorrect to use it in your start script and on Railway, so it is something you should remove from the start script and report back if the issue continues. It would also be a good idea to make sure Railway is indeed using the start script, you can see what start command Railway will use in the build table at the top of your build logs.

Alright! Will make the app a little more production-ready, and check if the issue persists. If so, I'll report back. Thanks for now!

Sounds good!

Hey, update on this ticket. Did a test for some hours now, seems that on 1000 requests, 3 fail. Which is nothing, but still raises the same question as in the beginning. Any clue if this could be on Railways side, or a wrong configuration?

Are you able to consistently reproduce that?

No unfortunately not. It seems very occasionally. The other time I tested it, I got no errors at all

Whats the memory usage like? You only have 500MB available on the trial plan.

Overview of the metrics attached

Would it be possible to ask you to test if you can reproduce this behaviour with the express hello world code?