Description

Normally when a containerd-shim exits it is handled by terminating the process that was running (assuming it was running).
If containerd is restarted (or most likely if containerd is not the parent process of containerd-shim) the exit is not handled.

When this happens ctr t ls does not show the task corresponding to the shim.
If you attempt to start a new task you get an error like:

ctr: mkdir /var/run/docker/containerd/daemon/io.containerd.runtime.v1.linux/moby/c548dfcd2ba36c863b5a8dc194fb01c82a4de715fb56582a2ab2c3c18ab1d254: file exists: already exists

Steps to reproduce the issue:

1. Create container + start task
2. restart containerd
3. kill (either SIGKILL or SIGTERM) containerd-shim

Describe the results you received:

• containerd-shim is not reaped
• containerd-shim socket is left behind
• container is unusable, must delete and recreate.

Describe the results you expected:

containerd should be able to reap shim resources

Output of containerd --version:

containerd github.com/containerd/containerd v1.0.1 9b55aab90508bd389d7654c4baf173a981477d55