#793516 killer kills running sshd session instances

#793516#5
Date:
2015-07-24 19:22:32 UTC
From:
To:
Today I saw killer kill sshd user processes (i.e., active SSH sessions).

This should not happen. It happened on a Debian Edu jessie main server (after (dist-)upgrade from squeeze).

I will investigate further and provide more info while upgrading our customer site / deploying a new school in Lübeck.

Mike

#793516#10
Date:
2015-07-24 19:52:21 UTC
From:
To:
[Mike Gabriel]

killer depend on updated utmp/wtmp information (check last/w/who).  If
the active SSH sessions did not register their login with utmp, killer
can not know their processes belong to active login sessions.

#793516#15
Date:
2015-07-24 21:24:31 UTC
From:
To:
Hi Petter,

hmmm... ok... will check that (the TJENER runs on systemd, which
should handle that IMHO).

Mike

#793516#20
Date:
2015-10-09 09:08:35 UTC
From:
To:
As an alternative, I'm just wondering if the killer job could be taken
over by systemd-logind, see 'man logind.conf'. This would allow to keep
processes of stated users and other actions, too.

Also, at least with systemd version 226, it would be possible to drop
configuration snippets into /lib/systemd/*.conf.d/.

Wolfgang

#793516#25
Date:
2015-10-09 11:44:15 UTC
From:
To:
[Wolfgang Schweer]

Could be.  Its task is to kill leftover processes, meaning non-system
and non-niced processes belonging to no longer logged in users.  And the
problem we are having is identifying logged in users, as the method to
do this seem to be in flux.

#793516#30
Date:
2015-10-09 13:00:59 UTC
From:
To:
I have to remove killer fromm all ws.
Otherwise users get logged out every 30 min

Regards

Giorgio

Il ven ott 9 13:44:15 2015 GMT+0200, Petter Reinholdtsen scrive: