Large container pushes to GHCR in parallel are timing out after 6 hours

For about the last week or so, pushes have been timing out sporadically. Is there a service issue on the receiving end of the registry?

One example of the error:

error: failed to copy: failed to do request: Put "ghcr.io/v2/integer32llc/rust-playground-ci-rust-nightly/blobs/upload/6efdd15f-c4d7-4421-ab7e-2d7760e77633?digest=sha256%3Aee7f822a08e47dc7726f8f688824245dae84d955ca13a73b4b4f66219ef78752": write tcp 172.17.0.2:54112->140.82.113.33:443: write: connection reset by peer

I also have this problem, any solution? I have to retry several times before the job is successful. Please take a look here. Thanks

I’ve also been following Pushing a multi-platform image to ghcr.io results in an endless loop · Issue #834 · docker/buildx · GitHub, which suggests pinning buildkit to 0.9.1:

uses: docker/setup-buildx-action@v1
with:
  driver-opts: image=moby/buildkit:v0.9.1

I’ve tried this on my repo. Although there haven’t been many attempts since then, they’ve all been successful.