race: completion: cannot get cgroup path unless container is running

[edsantiago]

Another parallel-system-test flake, seen in f39 root:

<+036ms> # # podman pod stats --no-stream --format {{"\n"}}
<+198ms> # Error: cannot get cgroup path unless container e3c650c8438160f0c2728fc4ed589bc6c06546e6abfba38db49d5d15b9325310 is running: container is stopped
<+016ms> # [ rc=125 ]
         # #/vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
         # #|     FAIL: podman pod stats --format '{{"\n"}}'
         # #| expected: = ''
         # #|   actual:   Error: cannot get cgroup path unless container e3c650c8438160f0c2728fc4ed589bc6c06546e6abfba38db49d5d15b9325310 is running: container is stopped
         # #\^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

[Luap99]

Reproducer
terminal 1

while :; do bin/podman run --rm -d quay.io/libpod/testimage:20240123 sleep 1; sleep 0.1; done

terminal 2

$ bin/podman stats --interval 1
<--shows stats for a bit and then fails with
Error: cannot get cgroup path unless container xxx is running: container is stopped

Likely another case of must ignore the "container is stopped" error when looping over all containers.

[Luap99]

The reproducer failed quickly for me, with my PR it looks stable now (running for 15 mins)

[edsantiago]

Sorry, still happening

[Luap99]

Oh sorry, I was looking at podman stats all this time, podman pod stats works slightly different and doesn't make use of the "all" option which means the error is not ignored there.

I still fixed a valid issue with podman stats at least but pod stats uses a different code path...