sean/binjovi
Watch
1
0
Fork
You've already forked binjovi
 0
Code Issues Pull requests 1 Projects Releases 745 Packages Wiki Activity

fix(deploy): pin binjovi-flux-deployer v11, and drop the archived repository from the chain #1792

Merged
binjovi-bot merged 1 commit from fix/chain-archived-and-deployer-v11 into trunk 2026-09-14 20:28:49 +00:00
Conversation 0 Commits 1 Files changed 3 +10 -7
sean commented 2026-09-14 20:26:54 +00:00
Owner
Copy link

What

  • Pin binjovi-flux-deployer v11 (sha256:dbddca66…, multi-arch) in both FROM lines and container_contract_test.exs. v11 (sean/builder-images#365) lets the post-push observation of Codeberg main outlast Codeberg's read lag: 30 reads 3 s apart instead of 5, with a failure that names the read count, push status and last value seen. Root cause of every seanfarm deploy today failing its first attempt ~16 s in and passing on retry-deploy.
  • Drop pipelines-operator from the chain table. It is archived on the cluster; Forgejo refuses a push mirror on an archived repository, so both of its creates failed closed (denied) on v0.1.742 and the heal would retry them every 30 min for ever. The comment records that it has no copy off the cluster and how to give it one (unarchive for one converge).

Verified

  • container_contract_test.exs: 13 passed. forgejo_push_mirror_bindings_test.exs: 4 passed. mix compile --warnings-as-errors, mix format --check-formatted, make contract-check clean.
  • Chain state after v0.1.742 (Forgejo side): hop 1 present on 45/46, hop 2 present on 45/46 — the one absence on each hop is pipelines-operator; the one last_error (rpg-munchers-tng, "reference already exists") was a first-push race and both copies already hold identical refs.

https://claude.ai/code/session_01XdBRc9CHgFhAxCyHbdVtt6

## What - **Pin `binjovi-flux-deployer` v11** (`sha256:dbddca66…`, multi-arch) in both `FROM` lines and `container_contract_test.exs`. v11 (sean/builder-images#365) lets the post-push observation of Codeberg `main` outlast Codeberg's read lag: 30 reads 3 s apart instead of 5, with a failure that names the read count, push status and last value seen. Root cause of every seanfarm deploy today failing its first attempt ~16 s in and passing on `retry-deploy`. - **Drop `pipelines-operator` from the chain table.** It is archived on the cluster; Forgejo refuses a push mirror on an archived repository, so both of its creates failed closed (`denied`) on v0.1.742 and the heal would retry them every 30 min for ever. The comment records that it has no copy off the cluster and how to give it one (unarchive for one converge). ## Verified - `container_contract_test.exs`: 13 passed. `forgejo_push_mirror_bindings_test.exs`: 4 passed. `mix compile --warnings-as-errors`, `mix format --check-formatted`, `make contract-check` clean. - Chain state after v0.1.742 (Forgejo side): hop 1 present on 45/46, hop 2 present on 45/46 — the one absence on each hop is `pipelines-operator`; the one `last_error` (`rpg-munchers-tng`, "reference already exists") was a first-push race and both copies already hold identical refs. https://claude.ai/code/session_01XdBRc9CHgFhAxCyHbdVtt6
fix(deploy): pin binjovi-flux-deployer v11, and drop the archived repository from the chain
All checks were successful
binjovi/ci Binjovi completed the frozen plan
Details
bb252aa8b1
 
Two facts from the first hours of v0.1.742.

1. Every seanfarm deploy today failed its first attempt and passed on
   retry-deploy. The native task exited 1 about sixteen seconds after it was
   claimed: the push to Codeberg main had returned 0, and the deployer's five
   reads of Codeberg main three seconds apart never saw it. Codeberg's read
   path lags its write path. builder-images v11 of binjovi-flux-deployer
   (sean/builder-images#365) keeps the independent observation but lets it
   outlast the lag: thirty reads, three seconds apart, and a failure that
   names the count, the push status and the last value seen. This pins it in
   both FROM lines and in the container contract.

2. pipelines-operator is archived on the cluster. Forgejo refuses a push
   mirror on an archived repository, so both of its chain creates failed
   closed (`denied`) and the heal would have retried them every thirty
   minutes for ever. It leaves the table; the comment records that it has no
   copy off the cluster at all and how to give it one.

Claude-Session: https://claude.ai/code/session_01XdBRc9CHgFhAxCyHbdVtt6
binjovi-bot deleted branch fix/chain-archived-and-deployer-v11 2026-09-14 20:28:49 +00:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
sean/binjovi!1792
No description provided.