S3 storage mode - startup message doesn't indicate that S3 mode is enabled

[kpfleming]

I've built from the main branch, and I've setup a YAML configuration file that looks like this:

listen: "127.0.0.1:8081"
base_url: "https://pkg-proxy.km6g.us"

storage:
  url: "s3://pkg-proxy?awssdk=v2&endpoint=[::1]:3910&use_path_style=true&disable_https=true&region=km6g"
  max_size: "20GB"

database:
  driver: "sqlite"
  path: "/var/lib/pkg-proxy/cache.db"

log:
  level: "info"
  format: "text"

If I modify the storage.url setting to make it invalid (a misspelled query parameter), then at startup I'll get an error message and the proxy doesn't start. This is expected.

If I correct the errors and provide a valid URL, then startup succeeds, but the proxy reports this:

time=2026-03-31T20:12:34.526-04:00 level=INFO msg="starting server" listen=127.0.0.1:8081 base_url=https://pkg-proxy.km6g.us storage=./cache/artifacts database=/var/lib/pkg-proxy/cache.db

I also don't see any connection attempts in the log of my S3-compatible server (which is Garage).

[kpfleming]

I went further with my configuration and actually pulled some content through the proxy, and it was stored in the S3 object store as expected. I guess this means that the connection to the S3 store is made 'on demand' and not kept open, so while this is all working fine the startup message is a bit misleading :-)

[andrew]

Yeah that messaging could definitely be better, i'll call that a bug for now.

[kpfleming]

An additional improvement could be issuing a HeadBucket operation during startup to validate that the supplied S3 URL (and credentials in environment variables) are valid and usable. Without that, there's no way to know that they are valid and usable until a proxy request arrives and an attempt is made to use the blob store.