Samba 4.10 and later versions have a process restart capability to greatly reduce the impact of crashes due to a NULL pointer de-reference or abort(). However SIGKILL was deliberatly omitted. Sadly this is the most likely case, due to the OOM killer, as raised here: https://lists.samba.org/archive/samba-technical/2019-November/134529.html Subsequent discussion (offline) has been to agree that we should restart in this case.
i had smbd killed by oom due to a bug and found no smbd processes running afterwards. isn't that systemd's job to restart failed services ?
Fixed in Samba 4.12.0 with commit 30cbc5b90e4951a0ea924c204ff73e15c9750ace Author: Andrew Bartlett <abartlet@samba.org> Date: Thu Dec 19 11:20:30 2019 +1300 s4-smbd: Also restart prefork children lost to SIGKILL (-9) Samba 4.10 and later versions have a process restart capability to greatly reduce the impact of crashes due to a NULL pointer de-reference or abort(). However SIGKILL was deliberatly omitted. Sadly this is the most likely case, due to the OOM killer, as raised here: https://lists.samba.org/archive/samba-technical/2019-November/134529.html Subsequent discussion (offline) has been to agree that we should restart in this case. BUG: https://bugzilla.samba.org/show_bug.cgi?id=14221 Signed-off-by: Andrew Bartlett <abartlet@samba.org> Reviewed-by: Ralph Boehme <slow@samba.org>
(In reply to roland from comment #1) Yes, if the smbd parent process dies, it is for systemd to restart it. This bug is about the "samba" process used for the AD DC, and the children of that process.