no error shown in the build logs?

Ah, yeah, sorry, you can't see my connected service in here:

It just says that it failed while starting the deployment.

I can confirm that I can docker run -it --rm kapono/tranquil-pds successfully after removing the image locally so it seems like nothing corrupted and the registry is serving the image fine.

The container starts with a log, message, even if you provide no environment variables at all.

I'm not using the official PDS at all.

You can see in the logs that the deployment is erroring before the container is even started with no feedback except to conact support, which seems to indicate a railway failure.

Are you just posting AI responses?

workdir "/app" does not exist on container

Something else is wrong here I think.

Deploying kapono/tranquil-pds:not-work-earlier ( kapono/tranquil-pds@sha256:c67226e166ec0edfdc3997b43c10f4b012304c22acb2278ad93afbe934c9d429 ) will timeout and fail creating the container.

But if I make a new image:

FROM kapono/tranquil-pds:not-work-earlier

RUN echo 'test'

Pushed to kapono/tranquil-pds:fix ( kapono/tranquil-pds@sha256:e3a0b4f7afbb64bdb2f233d6e612d1cc17ba6a4f519b347476214f61d46ec4a4 ), this meaningless change will fix the issue, without modifying the container filesystem, other than adding a new, blank layer.

It seems that the issue only happens if I docker pull directly from atcr.io, docker tag it, and then push it to DockerHub. If I add a new layer, and docker build it locally be fore pushing, it does work.

So there's something relatively subtle wrong with the way Railway is reading the atcr docker images, though Docker itself doesn't have a problem with it.

We don't use Docker.

Ah, yeah, cool, but is it expected that a valid docker image I can push to DockerHub should work?

Or at least have some feedback on what is wrong with my image so I can fix it or tell the atcr.io folks what is wrong with the image?

Something wrong with the set working directory would not make it an invalid image, so you could definitely still push it.

I already have a PR up waiting for CI to complete so that we fail fast with a The working directory "/app" does not exist in the container image. error.