Registry:3.0.0 failed with private s3 compatible bucket about distribution HOT 5 CLOSED

panic: s3aws: RequestError: send request failed
caused by: Get "https://docker-registry-test.some-s3-endpoint.net/?list-type=2&max-keys=1&prefix=scheduler-state.json": dial tcp: lookup docker-registry-test.some-s3-endpoint.net on 192.168.65.7:53: no such host
goroutine 1 [running]:

This is failing DNS lookup which has nothing to do with registry I'm afraid. You're gonna have to sort that out first.

As an aside it's strange that registry is panicking when the DNS lookup fails, though 🤔

from distribution.

@milosgajdos Thanks for your super fast reply!
My colleagues suspect this is DNS issue also, but it somehow worked well with registry:2.8.3, with exactly the same config for s3 in config.yml. Then it became a bit strange.
Do we have any official document for the configuration for registry:3.0.0? I am not sure if the format for the configuration file has changed and thus caused this.
I will also try to check the code if anything changed

from distribution.

The latest docs are here https://distribution.github.io/distribution/about/configuration/

from distribution.

I seem to find the issue:

In registry 3.0.0-alpha, you have this commit which allows independent setting of "ForcePathStyle" even when the RegionEndpoint is set:
8654a0e

In version 2 this was set to True by default when the RegionEndpoint Parameter was present:
https://github.com/distribution/distribution/blob/v2.8.3/registry/storage/driver/s3-aws/s3.go#L428

I have limited knowledge with s3 configurations, but since in my configuration the "forcepathstyle" option was not present, it went False and caused the errors.

Thanks for your quick replies again and I think this issue can be closed.

from distribution.

Awesome! I remember that PR and I completely forgot to think of that. Thanks, closing.

from distribution.