[Copilot]

The scheduled container build was failing in the Build nvidia (amd64) job during image publish, despite successful build/rechunk/tag stages. Root cause was an unhandled transient 500 Internal Server Error from GHCR during podman push.

• Failure mode addressed

  • Push To GHCR previously failed fast on the first registry-side error, aborting the entire workflow run.

• Workflow change

  • Added bounded retry logic to the GHCR push loop in .github/workflows/build.yml.
  • Each tag push now retries up to 3 times with incremental backoff before failing hard.
  • Preserved existing behavior for successful first-attempt pushes and digest output handling.

• Resulting behavior

  • Transient GHCR failures no longer fail the run immediately.
  • Persistent failures still fail deterministically with explicit error output.

max_attempts=3
for tag in ${ALIAS_TAGS}; do
  for attempt in $(seq 1 "${max_attempts}"); do
    if sudo -E podman push --digestfile=/tmp/digestfile \
      ${IMAGE_NAME}:${tag} ${IMAGE_REGISTRY}/${IMAGE_NAME}:${tag}; then
      break
    fi
    [[ "${attempt}" -eq "${max_attempts}" ]] && exit 1
    sleep $((attempt * 10))
  done
done

[Copilot]

Copilot wasn't able to review any files in this pull request. Check if the Files changed in this pull request are included in default exclusions.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.