Docker "device or resource busy" - Fix Unmount/Remove Failures in CI

Docker could not remove a container, volume, or path because the kernel says it is still in use. Something still has the mount or a file open, so the unlink/unmount is refused.

What this error means

A docker rm, docker volume rm, or container shutdown fails with device or resource busy (often unlinkat /var/lib/docker/...: device or resource busy). The resource lingers and can block re-creating it on the next run.

docker rm output

Error response from daemon: driver "overlay2" failed to remove root filesystem
for a1b2c3...: unlinkat /var/lib/docker/overlay2/.../merged: device or resource busy

Common causes

A mount inside the container is still active

A volume or bind mount that did not unmount cleanly keeps the container’s filesystem busy, so its root cannot be removed.

Another process holds an open file/handle

A leaked process, a still-attached docker exec session, or a host process with an open file under the container’s graph dir keeps it busy.

Leftover state on a reused runner

On a long-lived runner, a prior job left a half-cleaned container/volume whose mounts are still referenced.

How to fix it

Force-remove and prune leftover state

Stop and force-remove the container, then prune dangling volumes.

Terminal

docker rm -f <container> 2>/dev/null || true
docker volume rm -f <volume> 2>/dev/null || true
docker system prune -f

Find and release what holds the mount

Ensure no docker exec session is still attached to the container.
Check for host processes with open files under the container path and stop them.
As a last resort on an ephemeral runner, restart the Docker daemon to release stale mounts.

How to prevent it

Use --rm for ephemeral containers so they unmount and self-clean on exit.
Close exec sessions and stop child processes before removing containers.
Add a cleanup/prune step on long-lived runners between jobs.