4a776254-6ff4-46c3-b88c-07953cfddb2e

thank you all 🙂

COPY ./target/release/taille-auth /taille-auth

Yet target/release/taille-auth does not exist in that repo.

Its a good thing it doesn't though, you want to have the Dockerfile build the binary, not copy it in.

I don't want to use the dockerfile to build the binary, that's why it's done in the github workflow

is the cli copying the source code from the repo instead the . ?

https://github.com/railwayapp/cli/blob/master/src/commands/up.rs#L229-L239

it appear it does

hmmm

@Brody in that link, if you check the step before, I added a ls to check if target exists

Inside the dockerfile running in railway

but the file is there under target/release

I could upload the binary and download it on the dockerfile but it seems cumbersome

why do you not want the Dockerfile to build the binary?

because I dont want to waste cpu time on railway = cost

we don't charge for build?

and even if we did, i still could not recommend what you are attempting to do

any specific reason? both are on ubuntu-latest

its needless added complexity

ok, just find odd that the cli doesn't copy all the files in ., I checked the source code but I couldn't figure out why. I'd expect it to do something like docker build . and then run.

the cli does not run docker build

(I mean the integration)

the cli is probably only uploading the files to your server

you are also telling the cli not to copy the target folder -

ok, so railway uses the gitignore

because it's correct, I have target on my local machine and I dont want on remote

but it's generated when building the project in the workflow

I am modifying it to use dockerfile as you said

then you can forego the workflow all together, simply deploy via github

hmm, everything appears to be up and running but:

{
  "status": "error",
  "code": 502,
  "message": "Application failed to respond",
  "request_id": "Ka9mROjeTSi6yw-aVO1Z_A_499424464"
}

https://railway.com/project/4a776254-6ff4-46c3-b88c-07953cfddb2e/service/a2647a59-54ec-4b5b-87fd-0bacc2b5a535?environmentId=e695c3d3-cb29-460b-9fcf-34c28e4989bc&id=5876d22a-bdb6-4206-a735-2b7307f74a71#deploy

(I tried 0.0.0.0 and it didn't work)

but I will try again

hmm, it's working oddly

your target port was likely wrong

it has the address, 127.0.0.1:3000

but now it's always giving 404

404 is returned from your app

odd, it's stuck in Starting Container now

I only did changes in the server internally

just did a redeploy (through the railway UI) and listening now

😦 still, sometimes work, sometimes not

i would recommend you add more verbose logs to your app so that you can figure out why its sometimes failing to start

@Brody That's odd, running docker build -t taille-auth . and docker run -it -p 3000:3000 taille-auth in the Dockerfile from main branch works just fine and I can use the server in my localhost.

On Railway, the logs keep stuck on Starting Container

https://railway.com/project/4a776254-6ff4-46c3-b88c-07953cfddb2e/service/a2647a59-54ec-4b5b-87fd-0bacc2b5a535?environmentId=e695c3d3-cb29-460b-9fcf-34c28e4989bc&id=3bb41d7b-00e1-4c65-b84c-3fb88eec2b58#deploy

I don't think the server is crashing at all

nothing yet, I tried adding a healthcheck, removed tini, added more logs, nothing 😦

it's still working on my localhost

your current deployment has exited

yes because the healthcheck killed it

I am out of options, thinking in manually deploy my server on AWS instead

not sure what to tell you but this isn't an issue with the platform, when you only see starting container, your app has already exited.

you are correct, sorry for being rude with you

not being rude at all

@Brody I believe I found the issue, it's because inside the docker I was applying some optimisation to compile the dependencies in a layer up in the list, but it wasn't invalidating the binary. Then the new code was coming, being compiled but it wasn't generating a new binary.

that's an interesting find!

I read it's recommended to keep railway to decide the port, how I can setup it in the dockerfile? (in the expose part)

railway does not read from the expose keyword in your Dockerfile, your application simply needs to listen on the PORT variable

it appears to be listening

my server is listening the port from the env

@Brody maybe I found a issue with that on the UI

if you forget to set it to automatic, there is no warning about that

shouldn't the port from the env to be 3000?

but it triggered 8080 on my server

it's working now if I set it to automatically set the port, just think a warning about that would make things clear

8080 is what Railway sets the variable to

ok, but I had 3000 manually set, shouldn't it use too?

no, the target port on the domain does not control the PORT variable

ok

so all good now?

all good 🙂

!s