ARC runner "Init:Error" (dind init) in CI

By Daniel Zoghalchali·Latchkey

An ARC runner pod uses init containers to prepare the runner, including the Docker-in-Docker setup in dind mode. Init:Error means one of those init steps exited non-zero, so the runner container never starts.

What this error means

The runner pod shows STATUS Init:Error or Init:CrashLoopBackOff. The main runner container stays uninitialized while the init step keeps failing.

$ kubectl get pod -n arc-runners arc-runner-set-xxxx
NAME                   READY   STATUS       RESTARTS
arc-runner-set-xxxx    0/2     Init:Error   3

Common causes

The dind or externals init step failed

In dind mode an init container copies the runner externals or prepares Docker; a missing volume, image, or permission makes it exit with an error.

Wrong containerMode configuration

A containerMode misconfiguration (for example dind without the matching volumes) leaves the init container unable to complete.

How to fix it

Read the init container logs

1. List the pod containers to find the failing init container name.
2. Read that init container logs for the exit reason.
3. Fix the referenced image, volume, or permission it names.

kubectl logs -n arc-runners <runner-pod> -c init-dind-externals

Use the supported containerMode config

Set containerMode to dind (or kubernetes) exactly as documented so the init containers and volumes are generated correctly.

containerMode:
  type: "dind"

How to prevent it

• Configure containerMode from the documented example, not by hand.
• Ensure init container images are pullable in the cluster.
• Keep runner and dind image versions compatible.

Frequently asked questions

Related guides

References

• GitHub Actions documentation