On Sun, 12 Mar 2000, Billy Harvey wrote:
> > > > Problem is that one's thing junk is other person's valuable thing.
> > > ...
> > > > Bottom line is that I don't belive any kind of AI in OMM will do the
> > > > right job.
> > >
> > > It beats killing totally random stuff like we do now.
> >
> > > Can't you agree that the OOM killer, while not perfect,
> > > is better than the random killing we currently have?
> >
> > Yes, but it still believe it remains a hack, not optimal solution.
> > .. and as they say 'hack last longest' once they get in.
> >
> > I belive, ideally *user* should have some way to decide what should die.
>
> How difficult would it be to extend the OOM grim-reaper to allow a
> user configuration file, perhaps with specifiable kill-first processes
> and kill-last processes? For example, one poster had stated that some
> users might prefer that their httpd processes were given the greatest
> tolerance, or another that his math crunching program be the omega.
> Either (or both) could be specified through a configuration file that
> allowed regular expression parsing, or parsing similar to the
> hosts.allow file.
At the time, I suggested moving the grim-reaper code into a userspace
daemon, to allow much greater/easier configurability. I can think of a few
examples it would be rather messy to include in the kernel; first, I would
kill off Netscape processes (and a few others), then more user processes,
and then (once all the user processes have gone) resort to dropping
daemons. As an ultimate resort, reboot (it could have been a kernel memory
leak).
For the time being, though, the simple "kill the most likely suspect"
patch should be a pretty good approximation.
James.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 21:00:23 EST