Unable to create 2+ VMs at the same time using XenServer

Attempts to kill process are futile:

http://paste.openstack.org/show/1603/

This sounds rather bad -- can someone else using XenServer confirm this ?

Update with kernel messages:

http://paste.openstack.org/show/1628/

I can build a number of instances at once with XenServer. I had a merge to fix an issue with parallel xapi calls maybe a month or 2 ago, and after then everything's been fine.

This is some sort of kernel bug according to the pastes above. Notice this line:

"
 kernel:[50922.735291] general protection fault: 0000 [#1] SMP
"

Due to the crash caught in the kernel, it has left some processes in a infinite disk wait state and thus they can't be killed. Since the kernel is hosed, all you can do is reboot.

What version of XenServer is this? What kernel?

Latest XenServer 5.6 SP2 running Debian Squeeze in dom0:

root@compute-10-13-1-39:~# uname -a
Linux compute-10-13-1-39 2.6.32-5-amd64 #1 SMP Mon Mar 7 21:35:22 UTC 2011 x86_64 GNU/Linux

Brian, presumably you don't mean Squeeze in dom0. Do you mean squeeze in domU?

If you are using unmodified XenServer and you're getting that kernel trace, then that's something that we will look at immediately. Please reproduce the bug, then run xen-bugtool in dom0, or use XenCenter > Tools > Get Server Status Report. I will escalate immediately to XS Engineering.

Feel free to send the bug report directly to me, if this is not the right forum for it.

Thanks.

Yup, I meant domU. I'll try to get a report to you as soon as I can. Thanks for the assistance.

I wasn't certain what to collect...so I collected everything. Let me know if I can help out, turn on debugging, or do anything else that might help expose the error! Thanks Ewan.

I haven't seen issues running multiple builds but I'm also using --xenapi_inject_image=false and dropping the image directly to the XenServer host instead of injecting through the domU. I'd suspect the builds could be colliding somehow when mounting the VHDs to the compute worker.

What's the status for this one ? Is it still happening with current code ?

I am unable to reproduce this in trunk.

Closing based on last comment -- please reopen if issue is still present.