Skip to content

Troubleshooting ​

Jobs stay queued.

  • Check the label: runs-on: pitlane must match the label shown in Repositories.
  • Check the target status in Repositories. error shows the GitHub error, often that the App is not installed on that repository.
  • All slots busy? The dashboard shows jobs waiting for a slot. Raise Parallel jobs.
  • waiting for memory: a slot is free but the job's predicted memory does not fit next to the running jobs (see Scheduling). It starts when one of them finishes. Add RAM, lower the Memory reserve, or turn memory admission off if the predictions are too high (the dashboard shows each slot's prediction next to its real use).
  • waiting for a strict slot: a strict job (public repository) waits for a slot reserved for strict jobs, an empty one, or a warm slot whose owner keeps another one, instead of wiping an owner's last warm cache. Raise Reserved slots for strict jobs, or set it to 0 to let strict jobs take any slot.

Jobs show no CPU or memory usage. Pitlane reads usage from cgroup v2, so some hosts have no data. Check with stat -fc %T /sys/fs/cgroup: cgroup2fs is v2, tmpfs is v1. Nothing is recorded on:

  • a host on cgroup v1;
  • Docker Desktop or rootless Docker;
  • Pitlane running in a container that does not see the host's /sys/fs/cgroup;
  • runners adopted after a Pitlane restart.

Jobs still run normally. See Resource usage per job.

A slot is in error. Usually privileged containers are not allowed. On Proxmox, the LXC needs nesting=1,keyctl=1. Check journalctl -u pitlane and docker logs pitlane-slot-0.

services: or container: jobs fail. Make sure you did not override DOCKER_HOST in the workflow. Service ports are reachable on localhost.

Disk fills up. Lower Build cache kept per slot, or reduce the number of slots. To reset a slot's cache entirely: docker rm -f pitlane-slot-N && docker volume rm pitlane-slot-N-docker, then restart Pitlane.

The runner version is too old. GitHub refuses runners that are too old. Upgrade Pitlane: each release pins a recent runner image. With a custom runner image, rebuild it on a recent ghcr.io/actions/actions-runner and update Runner image in the settings.

Sign-in fails with a redirect error. The Public URL must match the GitHub App's callback URL (<base url>/auth/callback). Edit it in the App settings on GitHub if you changed the address.

Logs. journalctl -u pitlane -f (systemd) or docker compose logs -f pitlane. Use PITLANE_LOG_LEVEL=debug for more detail.

Multi-architecture builds (docker/setup-qemu-action) fail with exec format error. Pitlane mounts binfmt_misc and registers the QEMU emulators when it starts (this is needed in unprivileged LXC containers, where nothing else keeps binfmt_misc mounted). If the log says binfmt_misc not mountable, Pitlane runs in a container that may not mount it (e.g. Docker Compose without extra privileges): run it with systemd on the host, or keep multi-arch jobs on GitHub-hosted runners.

Windows or macOS jobs. Pitlane only provides Linux runners; keep windows-latest / macos-latest jobs on GitHub-hosted runners.

Released under the MIT License.