[ready-to-test] PR #61: hermes_infra unmerge=unmerge (restart recovery) #62

Open
opened 2026-09-11 20:59:49 +02:00 by hermes · 1 comment
Contributor

Status

  • PR open, CI green
  • Ready to test
  • Staging deploy
  • Prod deploy
  • Closed

What

Tracks testing of admin/brencher#61.

PR

  • URL: #61
  • Branch: fix/hermes_infra-unmerge-recovery (head 04069fe5616b)
  • Diff: +1/-1 in backend/configs/hermes_infra.py
    • SharedStateHolderInMemory(unmerge=None) → SharedStateHolderInMemory(unmerge=unmerge)

Why

After every brencher container restart, the hermes_infra env failed
to recover through unmerge.progress() because the in-memory state
holder was constructed with unmerge=None. Result: every processing
cycle logged Empty branches set for both CheckoutMerged and
DockerSwarmDeploy. Other 6 configs use unmerge=unmerge and are
unaffected. PR brings hermes_infra into line with the rest.

Test plan

  1. Deploy the merged image to a single dev/staging instance.
  2. Wait ~15 min and grep logs for Empty branches set — expect 0
    new entries.
  3. Restart the brencher container and confirm the env recovers
    without UI interaction (state populated via unmerge.progress()
    in state.progress()).
  4. UI smoke test — open brencher UI, verify the hermes_infra branch
    row appears in the branches list (was empty pre-fix after restart).

Acceptance

  • Empty branches set log volume for hermes_infra drops to 0.
  • No new errors introduced in other envs.
  • Production restart of brencher-new no longer requires UI to clear
    the hermes_infra log spam.

Notes

  • Note: Gitea labels API requires write permission on the repo. Since
    the tracking bot only has pull access on admin/brencher, status
    is conveyed via the title prefix ([ready-to-test]) plus the
    checklist above. Update title and checklist as the PR moves through
    staging → prod.
## Status - [x] PR open, CI green - [x] Ready to test - [ ] Staging deploy - [ ] Prod deploy - [ ] Closed ## What Tracks testing of admin/brencher#61. ## PR - URL: https://gitea.rudolf.keenetic.link/admin/brencher/pulls/61 - Branch: `fix/hermes_infra-unmerge-recovery` (head `04069fe5616b`) - Diff: `+1/-1` in `backend/configs/hermes_infra.py` - `SharedStateHolderInMemory(unmerge=None)` → `SharedStateHolderInMemory(unmerge=unmerge)` ## Why After every brencher container restart, the `hermes_infra` env failed to recover through `unmerge.progress()` because the in-memory state holder was constructed with `unmerge=None`. Result: every processing cycle logged `Empty branches set` for both `CheckoutMerged` and `DockerSwarmDeploy`. Other 6 configs use `unmerge=unmerge` and are unaffected. PR brings `hermes_infra` into line with the rest. ## Test plan 1. Deploy the merged image to a single dev/staging instance. 2. Wait ~15 min and grep logs for `Empty branches set` — expect 0 new entries. 3. Restart the brencher container and confirm the env recovers without UI interaction (state populated via `unmerge.progress()` in `state.progress()`). 4. UI smoke test — open brencher UI, verify the `hermes_infra` branch row appears in the branches list (was empty pre-fix after restart). ## Acceptance - `Empty branches set` log volume for `hermes_infra` drops to 0. - No new errors introduced in other envs. - Production restart of brencher-new no longer requires UI to clear the `hermes_infra` log spam. ## Notes - Note: Gitea labels API requires write permission on the repo. Since the tracking bot only has pull access on `admin/brencher`, status is conveyed via the title prefix (`[ready-to-test]`) plus the checklist above. Update title and checklist as the PR moves through staging → prod.
hermes changed title from [testing] PR #61: hermes_infra unmerge=unmerge (restart recovery) to [ready-to-test] PR #61: hermes_infra unmerge=unmerge (restart recovery) 2026-09-11 21:01:18 +02:00
Author
Contributor

test comment

test comment
Sign in to join this conversation.
No labels
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: admin/brencher#62