This repository was archived by the owner on Sep 17, 2026. It is now read-only.
Repository navigation
Drop the last two per-service .github directories - #24
Merged
Merged
Conversation
7321ee6 pruned these for the original seven services. forgectl (#12) and indexing-service (#10) arrived afterwards and nothing looked again, so 21 files across two directories survived. GitHub reads workflows, dependabot config and CODEOWNERS only at the repository root, so none of them has ever run here -- but they describe a per-repo release and deploy flow that does not apply, which is exactly what someone reads and believes while building Phase 1. Deleted rather than kept as reference: indexing-service's releaser.yml, tagpush.yml, release-binaries.yml, release-check.yml, publish-ghcr.yml, deploy.yml and terraform.yml are the closest thing we have to a worked example of the release flow Phase 1 must build, but a stale copy in the tree is worse than a pointer. They are one `git show` away here, and still live in the polyrepo. **Found while checking, and it is not cosmetic: forgectl is the only service with genuine scheduled workflows.** metrics-faults.yaml (every 12 hours) and metrics-payments.yaml (every 30 minutes) run `forgectl metrics faults` and `forgectl metrics payments` against `environment: mainnet`, pushing to an OTLP endpoint. They still run today, in fil-forge/forgectl, because that repository still exists -- the copies removed here never did. So nothing stops right now, but **archiving fil-forge/forgectl stops mainnet fault and payment metrics** unless the monorepo picks them up first, and doing that needs a mainnet environment and two secrets configured here, which is a person's job. Recorded in MONOREPO_TODO.md and on the wiki rather than left in this message. No equivalent hazard among the seven pruned earlier: the only `schedule:` keys in those were dependabot intervals, checked across every file 7321ee6 deleted. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CGAGAib517Ae1kg8SCdcEt
Peeja
pushed a commit
that referenced
this pull request
Sep 17, 2026
…deadline #23 re-homes the CI wall-clock measurement that would otherwise live only on closed #21, plus two findings from reading main: Phase 1 has no release.yml to use, and tags are gated on the rename because a submodule tag must sit in the repository its module path names. #24 removes the last two per-service .github directories. #25 layer-caches the image builds in itest and e2e while images.yml stays cold as the canary -- the only one of the five open PRs that can fail. The hazard is on Needs Human Work because it needs a person: forgectl's metrics-payments and metrics-faults run against environment: mainnet every 30 minutes and 12 hours, pushing to an OTLP endpoint from repository secrets. They run in the polyrepo today. Archiving fil-forge/forgectl stops them silently, and reproducing them here needs an environment and secrets configured on this repository. Also records what was deliberately not attempted pre-rename and why: the release workflow (unverifiable until tags can be cut, and an unrunnable workflow is the silent-green shape this repo keeps deleting), compat.yml (needs published images to test against), the tag scheme (long-lived, hard to reverse, a recommendation rather than an agent's decision), and building the Dockerfile.release files (goreleaser-shaped, they need the release flow first). Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CGAGAib517Ae1kg8SCdcEt
Peeja
pushed a commit
that referenced
this pull request
Sep 17, 2026
TestUploadAndRetrieve/filesystem has failed 2 of 40 e2e runs, and both are the same shape: compose up reports a dependency unhealthy, a different container each time (upload-1 on run 160, plc-1 on run 182), always the filesystem permutation and never s3. plc-1's log says connect ECONNREFUSED to its postgres, one restart, then the healthcheck deadline. That is a startup race, not a broken image. Run 160 is on main at ff2f794, two and a half hours before any of today's branches existed, which is what rules #25 out -- it changes how images are built, so the red looked exactly like a regression until the main failure was found. The two permutations are t.Parallel(), so ~40 containers start on one runner at once; both observed failures being filesystem is suggestive of something systematic, but two data points is not proof. One re-run queued on #25, the single re-run allowed for establishing that a failure is not the PR's. The honest fix is compose healthcheck ordering, not a retry. Also: main is 144b316 -- Petra merged #22, #23 and #24 while this was running. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CGAGAib517Ae1kg8SCdcEt
Peeja
pushed a commit
that referenced
this pull request
Sep 17, 2026
Petra merged #20, #17, #22, #19, #23 and #24; main is 144b316 and the open list is #25 and #26. Dropped the merged rows from both tables rather than letting them accumulate. #25's e2e red turned out to be the pre-existing filesystem flake, and the one re-run that cleared it was warm -- so it produced the measurement as well as the green: 8 images in 23s against a 7m34s baseline, and the e2e job from 13m10s to 5m29s. #26 puts that in MONOREPO_TODO.md with the caveats attached rather than the headline alone. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01CGAGAib517Ae1kg8SCdcEt
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
7321ee6apruned these for the original seven services.forgectl(#12) andindexing-service(#10) arrived afterwards and nothing looked again, so 21 files across two directories survived.None of them has ever run. But they describe a per-repo release and deploy flow that does not apply here — which is exactly what someone reads and believes while building Phase 1.
Deleted rather than kept as reference
indexing-service'sreleaser.yml,tagpush.yml,release-binaries.yml,release-check.yml,publish-ghcr.yml,deploy.ymlandterraform.ymlare the closest thing we have to a worked example of the release flow Phase 1 must build. A stale copy in the tree is still worse than a pointer: they are onegit showaway in this repository's history, and they still live in the polyrepo.forgectlis the only service with genuinely scheduled workflows.metrics-payments.yamlforgectl metrics payments --payer … --otlp-endpoint …metrics-faults.yamlforgectl metrics faults …Both against
environment: mainnet, pushing to an OTLP endpoint.They still run today, in
fil-forge/forgectl, because that repository still exists — the copies removed here never did. So nothing stops on merge. But archivingfil-forge/forgectlstops mainnet fault and payment metrics, silently, and reproducing them here needs amainnetenvironment and two secrets configured on this repository. That's a person's job and it should be sequenced with the archival, not after.Written up properly in
MONOREPO_TODO.md(separate PR) and on the wiki rather than left in a commit message.No equivalent hazard among the seven pruned earlier — I checked every file
7321ee6adeleted, and the onlyschedule:keys in them were dependabot intervals.🤖 Generated with Claude Code
https://claude.ai/code/session_01CGAGAib517Ae1kg8SCdcEt
Generated by Claude Code