Skip to content

fix(release-controller): resubmit failed GuestOS election 142125 - #2044

Merged
pietrodimarco-dfinity merged 1 commit into
mainfrom
fix/resubmit-guestos-election-142125
Jun 8, 2026
Merged

fix(release-controller): resubmit failed GuestOS election 142125#2044
pietrodimarco-dfinity merged 1 commit into
mainfrom
fix/resubmit-guestos-election-142125

Conversation

@pietrodimarco-dfinity

Copy link
Copy Markdown
Contributor

What

Add NNS proposal 142125 to the top-level ignored_proposals list in release-index.yaml so the release controller resubmits the GuestOS election for base b95f4a32b41798de115aac9298b51dd1662f1da5 (rc--2026-06-04_04-52).

Why

Proposal 142125 (ReviseElectedGuestosVersions) FAILED on execution:

[Registry] Cannot retire versions {"62e841d27ea12451f504e74c54843b7cbf93ca4d"},
because they are currently deployed to a subnet!

The proposal elects b95f4a32… and carries an auto-computed retire list (fa3e237…, f31310f…, c755244…, 62e841d…). 62e841d… (rc--2026-05-07_04-27) was still deployed to a subnet at execution time, so the registry trapped on the retirement and — because the proposal is atomic — b95f4a32… was never elected.

62e841d… ended up in the retire list because versions_to_unelect() derives the active-release window from the Prometheus > 10 nodes "active versions" query; the subnet still on 62e841d… had too few reporting nodes to register as active.

The sibling proposal 142127 (electing the canister-logging variant fb721da…) executed fine because non-base variants get no retire list.

Why it is now safe to resubmit

All 49 subnets have since been upgraded off 62e841d… (currently all on a47e5434…), so retiring it is now legal. The reconciler treats the FAILED proposal as "already submitted" and will not resubmit on its own; ignored_proposals forces it to forget 142125 and submit a fresh election on the next cycle (no redeploy required).

Follow-up

  • Once the replacement election proposal for b95f4a32… is submitted, remove the 142125 entry from ignored_proposals.

Notes

  • No code change; config only. YAML validated against release-index-schema.json (ignored_proposals is an integer array).

Proposal 142125 (electing GuestOS base
b95f4a32b41798de115aac9298b51dd1662f1da5 for rc--2026-06-04_04-52) FAILED
because its auto-computed retire list included
62e841d27ea12451f504e74c54843b7cbf93ca4d, which was still deployed to a
subnet at execution time, so the registry trapped on the retirement and
the new version was never elected.

All subnets have since been upgraded off 62e841d..., so retiring it is now
legal. The reconciler treats the FAILED proposal as already submitted and
will not resubmit on its own, so add 142125 to ignored_proposals to force
a fresh election proposal on the next cycle. The entry should be removed
once the replacement proposal has been submitted.
@pietrodimarco-dfinity
pietrodimarco-dfinity requested a review from a team as a code owner June 8, 2026 16:54
@pietrodimarco-dfinity
pietrodimarco-dfinity enabled auto-merge (squash) June 8, 2026 16:56
@pietrodimarco-dfinity
pietrodimarco-dfinity merged commit c51dc7d into main Jun 8, 2026
7 checks passed
@pietrodimarco-dfinity
pietrodimarco-dfinity deleted the fix/resubmit-guestos-election-142125 branch June 8, 2026 17:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants