feat(cluster): pin the Docker context the role drives per host - #21
Merged
Merged
Conversation
The role followed the host's current Docker context, which other software changes: on a Mac running its node in Colima with Docker Desktop also installed, launching Docker Desktop switches the current context to desktop-linux, and the next run would build a second, empty node there. github_runner_cluster_docker_context (default empty, keeping the current behaviour) names the context explicitly. docker_context.yml resolves it per host, failing before the host is touched if the context does not exist or DOCKER_HOST is also set, and publishes the cli_context for the community.docker modules and a DOCKER_CONTEXT environment for every docker CLI call, including the recovery and promotion playbooks. Tasks delegated to a Headscale host take that host's own value. The host's current context is never changed.
tests/lib/docker_context.sh gives the playbook runs a throwaway Docker CLI configuration whose current context points at a socket that does not exist, plus a named context for the real daemon, so any docker call that ignored github_runner_cluster_docker_context fails. The in_cluster mesh scenario (role and recovery playbook) and the Litestream test (delegated Headscale hosts and promotion) run that way, check the current context is unchanged afterwards, and a context_checks scenario checks the role refuses a missing context and one set alongside DOCKER_HOST before writing anything.
The pinned-context check after the playbook made run_role end on an if whose condition was false, so it returned success even when the role failed, and compose_faulty saw its expected refusal as a pass.
Mearman
marked this pull request as ready for review
September 26, 2026 09:56
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
🎉 This PR is included in version 1.4.0 🎉 The release is available on:
Your semantic-release bot 📦🚀 |
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.
The cluster role works out which Docker daemon to drive from the host's current Docker context (
docker context show), and every docker CLI call it makes follows that context too. That breaks on a host with more than one engine. On a Mac that runs its node in Colima and also has Docker Desktop installed, just launching Docker Desktop switches the current context todesktop-linux, and the next run of the role then finds no node there and builds a second, empty k3s node in Docker Desktop instead of managing the real one.This adds
github_runner_cluster_docker_context, a per-host variable. Empty (the default) keeps today's behaviour. When it is set:community.dockermodule call in the role passes it ascli_context;DOCKER_CONTEXTset to it:docker compose version, the in-cluster Headscale image pull anddocker execprobes and key reads, the Headscale API key creation, the hosted server'svolume inspect/rmandcompose exec, the promotion'scompose stop/rm/runandexec, and every command inrecover_in_cluster_mesh.yml;DOCKER_HOSTis also set;The resolution lives in
tasks/docker_context.yml, which the role and both playbooks include. Tasks delegated to a Headscale server or standby resolve against that host and read its own value from the inventory. The promotion's stop of the old server reads the value from the inventory without probing, since that host may be down.Tests:
tests/lib/docker_context.shgives the playbook runs a throwawayDOCKER_CONFIGwhose current context points at a socket that doesn't exist, plus a context named for the pin that points at the real daemon, so any call that ignored the pin fails. The in_cluster mesh scenario (role plus recovery playbook) and the Litestream test (delegated Headscale hosts plus promotion) now also run in that mode, and afterwards check the current context hasn't changed. A newcontext_checksscenario checks the two refusals happen before anything is written. There's also a unit test for the filter that builds the environment.I ran the branch's collection against a real fleet in
--check --diffmode, both with the variable unset and with it set to each host's current context, and got the same results as the released collection.