In some quick testing on my fork, https://github.com/tootsuite/mastodon/pull/6754 seems to break Google Cloud Storage used in S3 compatible mode. Should any config changes be necessary with the upgrade?
I haven't been able to test in isolation yet, just wanted to see if anyone else is in a position to verify this behaviour. I deployed a build a few commits past 2.3.3 accidentally at one point under my current config files, and started seeing a bunch of "invalid credentials" kind of errors connecting to storage. Reverting this single commit and going back to 5.1 resolves that completely.
Will this need a heads-up when it hits a tagged release? Can anyone see if this happens on any other not-really-S3 storage providers?
cc @ykzts
S3_ENABLED=true
S3_BUCKET=example-mastodon-storage
AWS_ACCESS_KEY_ID=GOOGXXXXXXXXXXXXXXXX
AWS_SECRET_ACCESS_KEY=XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
S3_REGION=ap-northeast-1
S3_PROTOCOL=https
S3_HOSTNAME=storage.googleapis.com
S3_ENDPOINT=https://storage.googleapis.com
S3_SIGNATURE_VERSION=s3
S3_CLOUDFRONT_HOST=files.example.com
I used Google Cloud Storage. It works without problems.
Mine's pretty much identical but defined as multi-regional:
S3_ENABLED=true
S3_BUCKET=static.chinwag.org
AWS_ACCESS_KEY_ID=GOOGetcetc
AWS_SECRET_ACCESS_KEY=arealkeyhere
S3_REGION=us
S3_PROTOCOL=https
S3_HOSTNAME=storage.googleapis.com
S3_ENDPOINT=https://storage.googleapis.com
S3_CLOUDFRONT_HOST=static.chinwag.org
OK, I'm going to have to do some tests with a couple of fresh installs it looks like. As soon as I applied this commit, my storage broke instantly. Reverted it, everything's fine again. I'll see if I can get some time over the weekend to try to narrow this down.
@bremensaki any progress on this trouble shooting?
Unfortunately not, I got insanely busy. Hope to try the paperclip update patch in isolation to my test instance tomorrow.
Finally got a chance to pull a full 2.4.0 RC onto an instance using the same Google Cloud store and credentials and it's working.
I still have this issue with my real instance, but there's got to be something else about the environment that's triggering it. Not Mastodon's problem.
Most helpful comment
Mine's pretty much identical but defined as multi-regional:
OK, I'm going to have to do some tests with a couple of fresh installs it looks like. As soon as I applied this commit, my storage broke instantly. Reverted it, everything's fine again. I'll see if I can get some time over the weekend to try to narrow this down.