fix(kernel): pull hetzner-builder from the canonical registry #2156
Loading…
Reference in a new issue
No description provided.
Delete branch "fix/hetzner-builder-canonical-registry"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Every kernel release fails at the snapshot bake. Reproduced today on
linux-release-request-gqxn4: thebake-snapshotpod sits inImagePullBackOffoncodeberg.org/someara/hetzner-builder:v4@sha256:fea5546e...withNotFound.zot
s sync log is unambiguous:failed to get upstream image manifest details, error: manifest not found, repository: someara/hetzner-builder, reference: v4. The dailyregistry-package-prunerun keeps only the 12-hex build tags on codeberg, sov3andv4` were reaped. Since the 2026-09-02 registries.yaml incident the nodes carry no codeberg credential, so there is no fallback path either. The pin has been in place since 2026-07-30 and last worked for the seanfarm45 release on 08-19.code.sean.farm is canonical.
registry.sean.farm/build/hetzner-buildercarries the same lineage and is atv5, so this pins that by digest.build/*sits outside bothseanfarm-image-digest-pinnedand the Kyverno cosign rule, which scope toregistry.sean.farm/seanfarm/*, so no new admission surface.Note this is a v4 to v5 builder bump, not a like-for-like restore: the exact v4 digest is absent from both registries, so it cannot be recovered.
Found while enabling
CONFIG_BLK_DEV_DMfor the hcloud CSI driver (sean/linux#45). Separately,kubernetes/flux/infrastructure/node-snapshot-bakein sean/seanfarm still pins the codebergv3tag and has the same latent break.https://claude.ai/code/session_01MdSbMhzabSbpG8TtP9Ur3H