Getting strange errors in my traces "Error response from daemon: No such image: registry.dagger.io"

[median mesa]

Full text is

Error response from daemon: No such image: registry.dagger.io/engine:v0.15.3

exec docker inspect --type=image registry.dagger.io/engine:v0.15.3

failed to run command: exit status 1

Error response from daemon: No such image: registry.dagger.io/engine:v0.15.3

[surreal cave]

@median mesa that is simply the dagger CLI checking if the engine container is installed on your machine. If it doesn't find it, it will auto-install. You should never actually see that error - the fact that it gets surfaced to you is a bug. Does it actually bubble up the error all the way to the top? Or are you seeing it buried in a sub-span?

[median mesa]

The function completes without errors but it shows up at the bottom of the trace:

There's some other (maybe) unrelated errors showing up in some sub-spans:

[thin oracle]

hey @median mesa are you running Dagger in your local machine?

is the engine currently being provisioned with docker?

or do you have other container runtime installed?

[median mesa]

Hey @thin oracle I'm using both local and Github

Using docker locally though I also have podman running. Assuming it will default to using docker

[thin oracle]

@median mesa if you run docker image ls do you see the engine image?

[thin oracle]

Ok.. so you have the image locally. Do you still get the same error if you run the pipeline locally?

[median mesa]

So not the original Error response from daemon: No such image: registry.dagger.io/engine:v0.15.3 error message

But the weird remotes.docker.resolver.HTTPRequest error is common to both local and CI

[thin oracle]

Is the au.icr.io registry supposed to be public? From the screenshots it seems like the engine can't access that

If you run the pipeline with dagger call -vvvv do you have more info about the error?

[median mesa]

Trying now

Running dagger call -vvvv --base-image-name=erc-web-$(uuid) --image-tags=feature-daggerise build-publish --registry-hostname=ttl.sh

Using ttl.sh to rule out private registry issues

The trace doesn't show the errors though

[thin oracle]

Oh, seems like the image got pushed in this case?

So IIUC this seems to be a false positive, correct?

Does the pipeline succeed even though you get those red spans?

[median mesa]

Yep that's right!

Some weird issue with the publish() method?

[copper hornet]

I'm having a very similar issue, where it is complaining about 0.15.3 not being found in the registry.

huh, weird. I'm getting a dial TCP issue with
docker image pull debian
so... I'm going to look into this, maybe something with homebrew upgrade broke stuff for me today.

Weird. I updated Little Snitch and it seems to have broken things. I just disabled Little Snitch and its working fine.

[thin oracle]

I'm mostly sure those spans are coming from buildkit. IIRC someone also brought up this observation some time ago . Maybe @crimson trail recalls something about thay 🙏

[crimson trail]

yeah, just noise afaik. Probably just checking if each layer is already uploaded (HEAD), seeing that it's not (failed), and then uploading

[thin oracle]

wondering why buildkit would be marking those spans as failed thought? Maybe there's a generic check in the code that if the status !=2xx then it counts that request as failed and therefore, mark that span as error?

/me avoid 👀 at the buildkit code lol 😄