not ok 42 runc delete: cgroup not cleaned up correctly

[kolyshkin]

This is an interesting failure I haven't seen before. Happened in centos-stream-8 CI here: https://cirrus-ci.com/task/6376444287778816?logs=integration_systemd#L49

not ok 42 runc delete
# (from function `fail' in file tests/integration/helpers.bash, line 338,
#  in test file tests/integration/delete.bats, line 30)
#   `[ "$output" = "" ] || fail "cgroup not cleaned up correctly: $output"' failed
# runc spec (status=0):
# 
# runc run -d --console-socket /tmp/bats-run-68012/runc.IPOypI/tty/sock testbusyboxdelete (status=0):
# 
# runc state testbusyboxdelete (status=0):
# {
#   "ociVersion": "1.0.2-dev",
#   "id": "testbusyboxdelete",
#   "pid": 71979,
#   "status": "running",
#   "bundle": "/tmp/bats-run-68012/runc.IPOypI/bundle",
#   "rootfs": "/tmp/bats-run-68012/runc.IPOypI/bundle/rootfs",
#   "created": "2022-02-19T00:12:45.413050609Z",
#   "owner": ""
# }
# runc kill testbusyboxdelete KILL (status=0):
# 
# runc delete testbusyboxdelete (status=0):
# 
# runc state testbusyboxdelete (status=1):
# time="2022-02-19T00:12:45Z" level=error msg="container does not exist"
# cgroup not cleaned up correctly: /sys/fs/cgroup/pids/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount
# /sys/fs/cgroup/cpu,cpuacct/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount
# /sys/fs/cgroup/blkio/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount
# /sys/fs/cgroup/devices/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount
# /sys/fs/cgroup/memory/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount
# /sys/fs/cgroup/systemd/system.slice/tmp-bats\x2drun\x2d68012-runc.IPOypI-state-testbusyboxdelete-runc.zriC8C.mount

[kolyshkin]

Ah, OK, this seems like a cgroup created by systemd for the temporary mount unit, created by systemd as a result of runc doing the self bind-mount. Obviously, the test should not take this cgroup into account. IOW it's a flake, and the fix is something like:

diff --git a/tests/integration/delete.bats b/tests/integration/delete.bats
index ea8a3bb6..42253fd3 100644
--- a/tests/integration/delete.bats
+++ b/tests/integration/delete.bats
@@ -26,7 +26,7 @@ function teardown() {
        runc state testbusyboxdelete
        [ "$status" -ne 0 ]
 
-       output=$(find /sys/fs/cgroup -wholename '*testbusyboxdelete*' -type d)
+       output=$(find /sys/fs/cgroup -wholename '*testbusyboxdelete*' -type d | grep -v 'mount$')
        [ "$output" = "" ] || fail "cgroup not cleaned up correctly: $output"
 }
 

Yet, this is another reminder that we should try to implement something that avoids that mount entirely.