http://bugzilla.novell.com/show_bug.cgi?id=589788
http://bugzilla.novell.com/show_bug.cgi?id=589788#c9
--- Comment #9 from Jan Beulich 2010-03-24 10:39:50 UTC ---
Re #6: So there are (many) domains that failed to die completely. We're
observing a similar problem currently on SLE11 SP1 (Xen 4.0), but this issue is
still under investigation. But with that (and according to the provided output)
you should be seeing this with more than half of your guests. What's possibly
interesting is that while up to domain 100 no single domain got fully
destroyed, subsequently there are no zombie domains anymore (all domains with
IDs greater than 100 are still alive). Unless you have an explanation for this
(e.g. all properly destroyed VMs having been of another type, e.g. PV), this
might be a hint towards the root of the problem.
What I also note is that shadow-memory guests most of the time leak much more
pages than HAP ones (and - statistically meaningless - there's one PV zombie
with yet fewer pages left).
In Dom0, are you having more qemu-dm processes than there are guests?
Re #7: Actually, 'm' would probably be better than 'H' in case we'll need it
again.
Would you be able to rebuild the hypervisor if we provided you with a debugging
patch?
--
Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.