[libvirt-users] Kernel unresponsive after booting 700+ vm's on a single host

[Eric Blake]

On 09/10/2012 07:51 AM, Alfred Bratterud wrote:
> For a research project we are trying to boot a very large amount of tiny, custom built VM's on KVM/ubuntu. The maximum VM-count achieved was 1000, but with substantial slowness, and eventually kernel failure, while the cpu/memory loads were nowhere near limits. Where is the likely bottleneck? Any solutions, workarounds, hacks or dirty tricks?

Are you using cgroups?  There have been some known bottlenecks in the
kernel cgroup code, where it scales very miserably; and since libvirt
uses a different cgroup per VM by default when cgroups are enabled, that
might explain part of the problem.

Other than that, if you can profile the slowdowns, I'm sure people would
be interested in the results.

-- 
Eric Blake   eblake at redhat.com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 617 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/libvirt-users/attachments/20120910/23273202/attachment.sig>