Failure on GH but passing on local | Dagger | Page 1

frosty wharf Dec 12, 2024, 9:55 PM

#

Hello everybody!

I just upgraded from 0.13.7 to 0.15.0. It works just fine locally but when it runs on GH it fails. Interestingly enough, Dagger Cloud shows it as successful instead of a failure.

This is the trace when running on GH: https://dagger.cloud/fprotazio/traces/c5b54d31ca01e1653a974ebb13dda5ae

This is the trace when running locally: https://dagger.cloud/fprotazio/traces/1ef1ab219cee8be2640c821345c4a7eb#a2d3676ca1380df1

I attached the errors I am getting.

Could someone explain why it is failing?

crystal olive Dec 12, 2024, 11:09 PM

#

Hey!

This looks suspicious to me

could not find module entrypoint: class Barral from import. Class should be exported to benefit from all features.
missing export in class Barral at /src/.dagger/src/index.ts:5:0 but it's used by the module.

"./node_modules"

I am assuming that maybe you're caching node_modules somehow? If so, we don't recommend doing this because it can lead to issues like this

Instead of caching node_modules, we suggest caching the /root/.npm folder

#

I am confused about the image stuff to be honest, that makes no sense to me :/

I think I have seen this before, its very strange that it reports an exit 1 but then keeps going.

frosty wharf Dec 12, 2024, 11:17 PM

#

could not find module entrypoint: class Barral from import. Class should be exported to benefit from all features. missing export in class Barral at /src/.dagger/src/index.ts:5:0 but it's used by the module.

I got rid of that error by doing: export class Barral {}

crystal olive Dec 12, 2024, 11:18 PM

#

Yeah sorry I don't think that is the issue

frosty wharf Dec 12, 2024, 11:18 PM

#

Oh yeh, I am caching node_modules. Gonna change it, too.

crystal olive Dec 12, 2024, 11:19 PM

#

frosty wharf Oh yeh, I am caching node_modules. Gonna change it, too.

I think you should still do this, but I suspect there is something else weird happening here.

frosty wharf Dec 12, 2024, 11:20 PM

#

crystal olive I am confused about the image stuff to be honest, that makes no sense to me :/ ...

Me too. I was using engine/sdk 0.13.7 and everything was fine. I upgraded to engine/sdk 0.15.0 and I got no error on local from what I can tell but was failing on GH. Then, I upgraded to 0.15.1

#

It should be like that, right?

.withMountedCache("/root/.npm", dag.cacheVolume("npm"))

crystal olive Dec 12, 2024, 11:21 PM

#

frosty wharf It should be like that, right? `.withMountedCache("/root/.npm", dag.cacheVolum...

Yep!

crystal olive Dec 12, 2024, 11:22 PM

#

frosty wharf Me too. I was using engine/sdk 0.13.7 and everything was fine. I upgraded to eng...

I know we're in the middle of a relese at the moment it sometimes take a bit for the upgrade to happen everywhere. For example I am not able to brew upgrade dagger even though the CLI knows there is an update.

Your local trace seems to be using 0.15 though https://dagger.cloud/fprotazio/traces/1ef1ab219cee8be2640c821345c4a7eb#a2d3676ca1380df1

Do you have a local 0.15.1?

frosty wharf Dec 12, 2024, 11:24 PM

#

Yeh, all the images were taken with 0.15.0. Then, I realized there was a new version and upgraded again. So, yes, my local engine is running 0.15.1. My SDK is at the same version

#

#

crystal olive Dec 12, 2024, 11:27 PM

#

frosty wharf Yeh, all the images were taken with 0.15.0. Then, I realized there was a new ver...

I see can, you do me a favor and try to stop the engine locally (docker stop $YOUR_ENGINE_ID), then run your dagger command. I suspect this will cause the same error that you see on GHA.

frosty wharf Dec 12, 2024, 11:28 PM

#

okay, gonna try

crystal olive Dec 12, 2024, 11:28 PM

#

crystal olive I see can, you do me a favor and try to stop the engine locally (docker stop $YO...

Can you send me a the trace of that attempt?

#

The reason I am asking is that it looks like this command runs, but only when the engine is not already running (which is the case in GHA, but not locally)

[succeed on both]

exec docker ps -a --no-trunc --filter name=^/dagger-engine- --format {{.Names}}

[fail on cold start]

exec docker inspect --type=image registry.dagger.io/engine:v0.15.1

#

Hm, this may not be relevant, it seems that same failure occured even on older engines

i.e. https://dagger.cloud/fprotazio/traces/47c4ecf0d4a8c1a31f0dbe0aed83fb65

frosty wharf Dec 12, 2024, 11:31 PM

#

I ran docker stop <dagger-container-ID>.
I ran dagger call build --source=./

And this is the trace:
https://dagger.cloud/fprotazio/traces/b2dc2737d448efbaf97e893e6cb4621d

#

And this is the one I did a few minutes ago on GH after changing the cache location: https://dagger.cloud/fprotazio/traces/f79b17c2c905a004c6291e765c0cac11#d6f93e8fb596ba19

#

Errors from the GH one. Basically, the same ones.

crystal olive Dec 12, 2024, 11:44 PM

#

I am so surprised that this does not run the inspect command. Can you tell me more about your local docker setup?

frosty wharf Dec 12, 2024, 11:47 PM

#

Docker Engine is v27.4.0.
Docker Desktop is 4.37.0.

I enabled containerd image store. I guess that is the only unsual thing from the default instalation.

#

If I recall correctly, other dagger calls worked even after the containerd thing. But I can test with it disabled.

crystal olive Dec 12, 2024, 11:48 PM

#

Thanks so much for doing all the tests, I am trying to reproduce some stuff on my end too

frosty wharf Dec 12, 2024, 11:49 PM

#

no, no, thank YOU!

crystal olive Dec 12, 2024, 11:53 PM

#

I think all of this is a red herring

5   : │ │ │ │ exec docker ps -a --no-trunc --filter name=^/dagger-engine- --format {{.Names}} DONE [0.0s]
6   : │ │ │ │ exec docker inspect --type=image registry.dagger.io/engine:v0.15.1
6   : │ │ │ │ [0.0s] | []
6   : │ │ │ │ [0.0s] | Error response from daemon: No such image: registry.dagger.io/engine:v0.15.1
6   : │ │ │ │ exec docker inspect --type=image registry.dagger.io/engine:v0.15.1 ERROR [0.0s]
6   : │ │ │ │ ! failed to run command: exit status 1

My pipeline gets the same error but it does not cause an exit code 1 and keeps going. (failing for a different reason :D)

https://github.com/levlaz/snippetbox/actions/runs/12306813653/job/34349243054

Its still odd and we should fix that to remove the confusion, but it does not seem to be the core issue here.

GitHub

test gha · levlaz/snippetbox@bdc0fca

Contribute to levlaz/snippetbox development by creating an account on GitHub.

#

Can you share your full GHA output with me somehow>?

frosty wharf Dec 12, 2024, 11:55 PM

#

sure, one minute

frosty wharf Dec 12, 2024, 11:56 PM

#

crystal olive I think all of this is a red herring ``` 5 : │ │ │ │ exec docker ps -a --no-...

I agree. The image error is distracting. I deleted my local engine and run the dagger command and I got the image error too but it kept going and downloaded the latest image locally.

#

📎 logs_32009724823.zip

#

It fails in the Call Dagger Function step

#

Error: Unable to process file command 'output' successfully. Error: Invalid format ' "_type": "Container",'

Isnt the issu those two errors at the end of the Call Dagger Function step?

I dont understand why it shows those two errors only on GH.

crystal olive Dec 13, 2024, 12:11 AM

#

Can you share the block of the GHA config that is being called please?

frosty wharf Dec 13, 2024, 12:12 AM

#

Not sure I understand

crystal olive Dec 13, 2024, 12:12 AM

#

something like this

      - name: Run Dagger pipeline
        run: dagger call -m ci ci --dir . --env gha
        env:
           DAGGER_CLOUD_TOKEN: ${{ secrets.DAGGER_CLOUD_TOKEN }}

frosty wharf Dec 13, 2024, 12:14 AM

#

tame flax Dec 13, 2024, 2:53 AM

#

frosty wharf Hello everybody! I just upgraded from 0.13.7 to 0.15.0. It works just fine loc...

@frosty wharf @crystal olive I think I know what''s happening there. The pipeline itself is not failing, seems like the GHA error comes from trying to parse the output as it seems like GH action doesn't like that format it for some reason:

ref: https://github.com/dagger/dagger-for-github/blob/main/action.yml#L91-L100