docs: add RUNNERS.md, correct the stale review-runner claims - #81
Merged
Merged
Conversation
v3.4.0 changed claude-code-review.yml from the self-hosted pool default to ubuntu-24.04, and I did not update the README when I made that change. Three sections still told readers the opposite -- Runner strategy, Choosing a runner, and Self-hosted runners. Corrected, each noting the version the default changed in so the history stays legible. Adds RUNNERS.md for the consumer-side conventions the README does not cover, because they live in the consuming repos rather than here: the RUNNER_POOL and DIND_POOL variables, why they default in opposite directions, the flip and revert commands, the four jobs deliberately pinned to self-hosted, and the gotchas that cost time to find. It deliberately does not tabulate current state. An inventory of state held in another system goes stale -- a version-pin table in this fleet drifted three times in two days, twice while the PR carrying it was open -- so the doc gives the commands to read it instead. Those commands are verified: notably `gh search code` does NOT work for private repos here, returning an empty result rather than an error, so the doc warns against it and gives a workflow walk that does work. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0114gaSKj6YinQCkmxVbDACv
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Two things: a doc that did not exist, and three that my own change made wrong.
I left the README stale in #80
v3.4.0 changed
claude-code-review.ymlfrom the self-hosted default toubuntu-24.04. I did not update the README when I made that change, so three sections still told readers the opposite —## Runner strategy,### Choosing a runner, and## Self-hosted runnersall namedclaude-code-review.ymlas defaulting to self-hosted.Each is corrected and names the version the default changed in, so the history stays legible rather than silently rewritten.
claude.ymlgenuinely does still default to self-hosted and is unchanged.RUNNERS.md
The
RUNNER_POOLandDIND_POOLconventions were documented nowhere — only in comments inside the workflows that use them (visible once you are already editing the file you would need to change) and in merged PR bodies (effectively write-only). Nothing said which repos have which toggle, which way each defaults, what is deliberately pinned, or how to reverse any of it.The new doc covers:
RUNNER_POOLto self-hosted because the safe default is the status quo;DIND_POOLto hosted because there, hosted is the fix and defaulting to a single contended dind runner would preserve the faultmonitoring_stack's existing opt-out — so nobody "fixes" themubuntu-latest,ubuntu-22.04shipping Python 3.10, Dependabot PRs getting no secrets, the E2E concurrency guard that outlives its reason, and metered minutes not being poolable across orgsIt defers to the README for the
runs-oninput mechanics rather than restating them.It deliberately does not tabulate current state
An inventory of state that lives in another system goes stale. In this fleet a version-pin table drifted three times in two days, twice while the PR carrying it was open. So the doc gives the commands to read the truth instead.
Those commands are verified rather than plausible — and one finding is worth the warning it now carries:
gh search codedoes not index these private repos, and returns an empty result rather than an error — which reads as "nothing uses this". The doc warns against it and gives a workflow walk that actually works, tested againstsproncy(finds all 8 files across 5 repos).🤖 Generated with Claude Code
https://claude.ai/code/session_0114gaSKj6YinQCkmxVbDACv