v0.18.0: Optional "Webhook-based Autoscaling" to go beyond GitHub API rate limit, and various usability, reliability, and observability enhancements
CAUTION: If you're using the Helm chart, beware to review changes to CRDs and do manually upgrade CRDs! Helm installs CRDs only on installing a chart. It doesn't automatically upgrade CRDs. Otherwise you end up with troubles like #427, #467, and #468
This release includes the following changes from contributors. Thank you!
- Webhook-based autoscaling to go beyond GitHub API rate limit (#282, follow-ups #363, #381, #386, #390, #391, #395, #398, #404, $406, #407)
- Documentation available at https://github.com/summerwind/actions-runner-controller#faster-autoscaling-with-github-webhook
- Also thanks to @robwhitby for implementing the loadbalancer healthcheck endpoint (#378)
- @alex-mozejko has fixed the entrypoint script to gracefully handle self-updates (#284)
- TotalNumberOfQueuedAndInProgressWorkflowRuns has been fixed to support large-scale deployment (#295)
- actions-runner-controller is now able to automatically retry registration timeouts (#297, and #309 by @jonico)
- @jonico enhanced the controller to gracefully handle crashed nodes and stuck runner pods (#307)
- You can now add common runner labels per controller for e.g. easy "environment" labeling (#327)
- Runner registration check if now more reliable and has smaller chance of being rate limited by GitHub API (#405)
- @jonico fixed setting the runner group for enterprise runners (#376)
- actions-runner-controller is now able to restrict itself to a single namespace. This gives more chances to scale (#380, also #377 for related discussion and experience managing 100+ runners with actions-runer-controller)
- @tapih enhanced the controller support two or more RunnerDeployments in a namespace (#355, follow-up #386, #371)
- Rolling-update on RunnerDeployment has much smaller chance of disrupting progressing workflow jobs (#392)
- Various controller is now reliable- It has smaller chances of conflicts because it's updated to use Patch instead of Update whereas possible
- @int128 added metrics useful for observing the number of runners (#408)
- Runner controller is now part of our integration test suite. It should make subsequent releases more reliable (#409)
- Documentation has been greatly enhanced thanks to @callum-tait-pbx, and received various fixes from @mikeperry-kr, @ba32107, @robwhitby! (enhancements #303, #359 and #379, fixes #370, #373, #394, )
- Controller logs are now easier to read (#311 by @jonico, #410)
actions-runner-controller
is now listed in awesome-runners 🎉