Browse Source

Wrap stopping podman -t with systemd timeouts

Currently with KillMode=none, if podman -t hangs, the command repeats
leaving the process and its cgroup around.
But if a stop/start command hangs, we should not start another one.

Instead time it out properly via TimeoutStopSec set to the 2x of
the timeout given to the managed podman action. Then if it expires,
kill its cgroup all the way (KillMode=control-group is a default)
before rerunning the same operation. Also note that using KilMode
process is not recommended by systemd man pages.

Increase the grace stop timeout defaults 10->42s to align it with:
https://github.com/containers/podman/pull/8889

Closes-bug: #1945791
Change-Id: Iefe861f91cefe2a9cf773cae98b2440566ae8b5e
Signed-off-by: Bogdan Dobrelya <bdobreli@redhat.com>
(cherry picked from commit badd5a3c10f0af9e1c69a9b88064e743a36d4f53)
changes/20/814020/1
Bogdan Dobrelya 7 months ago
parent
commit
a783594fd4
  1. 3
      paunch/utils/systemd.py

3
paunch/utils/systemd.py

@ -83,6 +83,7 @@ def service_create(container, cconfig, sysdir=constants.SYSTEMD_DIR, log=None):
'wants': wants,
'restart': restart,
'stop_grace_period': stop_grace_period,
'kill_cgroup_period': 2 * stop_grace_period,
'sys_exec': '\n'.join(['%s=%s' % (x, y) for x, y in sys_exec.items()]),
}
# Ensure we don't have some trailing .requires directory and content for
@ -103,7 +104,7 @@ ExecReload=/usr/bin/podman kill --signal HUP %(name)s
ExecStop=/usr/bin/podman stop -t %(stop_grace_period)s %(name)s
ExecStopPost=/usr/bin/podman stop -t %(stop_grace_period)s %(name)s
SuccessExitStatus=137 142 143
KillMode=none
TimeoutStopSec=%(kill_cgroup_period)s
Type=forking
PIDFile=/var/run/%(name)s.pid
%(sys_exec)s

Loading…
Cancel
Save