[H-GEN] Systemd killing processes after logout

[Clinton Roy]

On 6 June 2016 at 10:24, Russell Stuart <russell-humbug at stuart.id.au> wrote:

>
> At the last Humbug we discussed in new feature in systemd - it now
> kills all background processes on logout, including those run by nohup,
> screen, tmux, mosh and so on.  This new behaviour arrived in Debian
> unstable on May 23.
>

In my mind, if you log out, all your processes should get killed. Things
like
tmux and screen that want to keep things hanging around should be required
to go through an extra hoop, and a hoop that's configurable by the sysadmin.
On a desktop system, I dare say that the default hoop (I think I just broke
the analogy)
should be for screen and tmux to `just work'.

Getting upstream to add code to jump through hoops..just isn't going to
work,
so I'm not sure how it's going to happen. At a guess there will be a
different, default
cgroup/session thingy that certain programs can get put or promoted into.


> At the meeting there was furious agreement this was insane, (except
> from Clinton, who defended it - but then he defended binary logs when
> they first came out so this wasn't unexpected).


I..don't remember it that way, of course. I really like the idea of
journalctl,
being able to quickly and easily search for a timestamp or a unit, or a
particular
error string in all your system logs is a great idea. Unfortunately the
implementation
has been subpar in my experience, slow enough to be basically useless,
and downright buggy (the latest one that was only fixed in latest rhel
..last week? was a compression
bug where systemd corrupted it's own logs). I've little doubt that speed
will
get better (well, it can hardly get worse) but I've stopped relying on
systemd journals
for everything but debugging purposes now.



> The question most
> interesting to me was WTF? - with the W standing for "why".  'nix has
> had a mechanism for do this for as long as I can remember - all
> descendent processes of the login process is sent a HUP signal when it
> exits.  It was simple, robust mechanism and had worked well back when
> we had 100's students learning to use computers for the first time on a
> single timesharing machine.  So why this new one?
>

I think here's where I disagree, asking processes to die on logout just
isn't the
right thing to do. Telling them they're going to die, then killing them,
is, imo.





-- 
Clinton Roy
Software Developer
Bloomberg VAULT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.humbug.org.au/pipermail/general/attachments/20160607/9e30aed5/attachment.html>