"Cannot pause container[uuid]: OCI runtime pause failed: unable to freeze: unknown"

[sk2]

Hi,
I am intermittently getting the above error when trying topologies such as the 5 node leaf-spine. I am using the latest ceos image.

I can run two nodes fine, but it seems once I get to 3-5 it seems to hit an issue. This makes me feel it is hitting a race condition where it tries to pause the container before it has fully been started.

Is this something others have hit before?

[networkop]

From my understanding the container.start() function should be blocking until the container transitions to the running state. This is the docker's source of this error:
https://github.com/moby/moby/blob/7b9275c0da707b030e62c96b679a976f31f929d3/daemon/pause.go#L41

So the container has definitely started since it's gone past this line.

I suspect it could be something to do with timeouts, e.g. not able to pause the init process in X seconds, but it looks like the error comes from several layers deep. Docker makes a call to containerd, which in turn calls runs, which is where it breaks.

Looking at the freeze function, it doesn't look promising:

 The numbers below are empirically chosen to have a decent
 chance to succeed in various scenarios ("runc pause/unpause
 with parallel runc exec" and "bare freeze/unfreeze on a very
 slow system"), tested on RHEL7 and Ubuntu 20.04 kernels.

Can you describe the system you're running on, e.g. CPU, OS version, Docker version?