[User Interview] railway up default behavior

[JakeCooper]

Suggestion

railway up - exit 0/1 for success/failed deploy
railway up --tail - holds for logs
railway up --detach - detaches immediately

Why

This allows CI to actually invoke the CLI and pass the correct exit code to the CLI to pass/fail CI

[alexng353]

Good. Also allow user to globally set default (railway config --global up.finish=exit|tail|detach)

[brody192]

I would like the defaults to stay as they are, I don't think its a good idea to change the default behavior, additional behavior should be a flag, with that said I would like a --wait or --report-ci flag, that waits for a the deploy to either fail or succeed without logs, then exits accordingly

[emilioschepis]

I'm also interested in this feature, I think it could be useful to have a good CI pipeline through services like GitHub Actions.

I don't know Rust that much, but I started working on a draft implementation for this. However looking at the subscription for build/deploy logs here I think that function will never return (it continuously waits on the next log).

Through the GraphiQL playground I also saw that there is no subscription for the deployment status, only a query (please correct me if I'm wrong) so I'm not really sure on how to approach this.

Thanks!

[brody192]

it would likely need to be achieved by polling for the deployment status, the same way that the dashboard does it

[maddsua]

That's literally what the cli currently does:

cli/src/commands/up.rs

Line 340 in 68db247

loop {

[maddsua]

I see a bigger problem in deciding which arguments set to choose. I'd like to modify initial proposal to look like this:

• railway up - wait until build finishes, stream build logs
• railway up --detach - leave it as it is (detach immediately after upload is complete)
• railway up --tail - stream all logs and only detach on build/critical runtime error (current default behavior)

[maddsua]

Guys I really wanna this resolved as it's one of those features that are not too hard to get done and it would greatly improve DX

[alexng353]

Guys I really wanna this resolved as it's one of those features that are not too hard to get done and it would greatly improve DX

lol cli development is pretty slow right now. If you guys come to a resolution for how you want to do it, I will make some time to PR some changes.

[maddsua]

One more idea: we could add a little notification about default behavior being changed right into the cli. Also add a lil counter to show that for the first five runs only

[maddsua]

Also, most CICD envs set an env variable and we could detect that

[maddsua]

Anyway, here it is: #477

[mattrossman]

This would be very helpful for our CI/CD pipeline on GitLab. We have followed the official guide "Using GitLab CI/CD with Railway" which recommends using the -d option on railway up, thus we always see jobs succeed in CI even when the build fails on Railway.

[maddsua]

#477 kinda solves this, we just need to decide on what to call the flag and we're basically done here

[mattrossman]

railway up --ci is short and sweet.

I'm coming from the Node.js ecosystem which has npm ci (technically "clean install" but doubles as "continuous integration" in practice) and jest --ci for precedent.

[maddsua]

Also can we please have a release so that I can finally use the damn docker image instead of running install script on each deploy? I could've just published the image myself, but having it not coming from railway will only complicate things down the road

[maddsua]

Guys, we can close this since it was resolved by #477 and #537