From: Jacques Gelinas (jack_at_solucorp.qc.ca)
Date: Fri 21 Mar 2003 - 17:07:15 GMT
On Thu, 20 Mar 2003 21:42:00 -0500, Tor Rune Skoglund wrote
> Dear Jacques, thank you for your reply,
>
> torsdag 20. mars 2003, 17:45, skrev Jacques Gelinas:
> > Now, what is missing to get a very very reliable solution
> >
> > -Resource limitation. If a vserver has a problem, it must not
> > bring the whole machine down.
> >
> > I have many of this stuff done (memory, file handle) but was
> > waiting for the issue with ctx-16 to be resolved.
> >
> > -Disk quota
> >
> > All this stuff is kind of around the corner and we were stocked with the
> > ctx-16 problem.
>
> Actually how far around the corner is it? (I know such things can be
> hard to predict... :)
>
> Do you suggest that for a server in a "production environment" I'll
> use ctx-13, or rather wait for the next patch with the additions
> you mention above?
Some people have been crashing on a regular basis with ctx-16. Also someone posted
a script crashing a box in no time. So I guess ctx-16-1 will be proven real soon.
wait a few days.
I will rename ctx-16-1 as ctx-17 so we will have a stable kernel. ctx-16 is adding
true BIND-any support compared to ctx-13. Better handling of UDP on loopback
as well.
Said differently, with ctx-16, there is less things to explain (less restrictions).
Right after ctx-17, I will release ctx-18 with per vserver file handle limitation
and per vserver memory limitation. The file handle stuff is done. The memory
still has a bug.
---------------------------------------------------------
Jacques Gelinas <jack_at_solucorp.qc.ca>
vserver: run general purpose virtual servers on one box, full speed!
http://www.solucorp.qc.ca/miscprj/s_context.hc