Package: openvpn
Version: 2.3.4-5
Severity: important
Hi,
Found the culprit. It seems to be start-stop-daemon which sends SIGKILL
too early while openvpn still tears down its stuff. I have noticed that
by stracing the app and seeing an unexpected SIGKILL.
After multiple attempts it seems that the tear down can easily be more
than 5s:
16835 18:56:53.657070 --- SIGTERM {si_signo=SIGTERM, si_code=SI_USER,
si_pid=16929, si_uid=0} ---
[...]
16835 18:56:58.941249 +++ exited with 0 +++
diff=5.284179
17013 18:57:15.537735 --- SIGTERM {si_signo=SIGTERM, si_code=SI_USER,
si_pid=17101, si_uid=0} ---
[...]
17013 18:57:21.257491 +++ exited with 0 +++
diff=5.719756
17187 18:57:40.792327 --- SIGTERM {si_signo=SIGTERM, si_code=SI_USER,
si_pid=17275, si_uid=0} ---
[...]
17187 18:57:46.247331 +++ exited with 0 +++
diff=5.455004
Maybe this is specific to my vpn configuration but increasing --retry to
10 has fixed my problem. Now it seems that pre-mature SIGKILL happened
much more often in the past. I was just lucky that the down callback was
called most of the time before the signal was triggered but the daemon
was still killed which is not the case anymore with the higher timeout.
Do you think the timeout can be increased by default or do I have to
update it only here because others do not see the same issue?
Thanks!
--
Michal Hocko
