segfault in nova.compute when deleting instances

asked 2014-10-29 04:54:00 -0500

pernacentus gravatar image

I'm running Icehouse atop Ubuntu server 12.04. I notice a very strange behaviour when I delete multiple vm instances. In fact, instances are kept in deleting state. By trying to analyze the problem, I found out that nova-compute, because of a segmentation fault error, was not running on the nodes containing the vms I was trying to delete. Here's the error from the log:

Oct 28 15:17:19 localhost kernel: [709293.487884] nova-compute[43959]: segfault at 0 ip 00007fe8d0da8c90 sp 00007fe8c4e55878 error 4 in libc-2.15.so[7fe8d0ca0000+1b5000

Is anyone experiencing this issue? How can i solve it?

Thanks in advance.

edit retag flag offensive close merge delete

Comments

I'm still having this issue. Again, is anyone experiencing this problem?

pernacentus gravatar imagepernacentus ( 2015-01-18 04:38:37 -0500 )edit

Where is running your compute-agent (in a VM)? Is there enough free memory ? Otherwise you can try to recompile your package. I've never see this issue before

foexle gravatar imagefoexle ( 2015-01-18 09:19:27 -0500 )edit

My compute agent is running on a physical server equipped with 2 16-cores AMD processors and 32GB of RAM. More than 20% of RAM was free when this error came out.

pernacentus gravatar imagepernacentus ( 2015-01-20 16:07:52 -0500 )edit

Could you paste more log entries (some lines before). In addition occurs that issue only on one host?

foexle gravatar imagefoexle ( 2015-01-22 03:28:48 -0500 )edit

No, it occurs on every compute node. Just to add more details, I deployed Openstack on each compute node by using Devstack. Here you can find more log entries related to nova: http://paste.openstack.org/show/160204/

pernacentus gravatar imagepernacentus ( 2015-01-22 05:03:36 -0500 )edit