Open Stack

Are you running glance over ssl?  We had an fairly random but consistent issue with vm's being stuck building.  Usually building means (at least what I found) the compute node did not have the backing image for the vm and needs to download it from glance , convert it from what ever backing file it was to a raw image, then start a qcow2 instances backed by the raw file.  Assuming you are the default configured raw backend qcow2.
You can check this by logging into the compute node and looking under: /var/nova/instances/_base/ for a <sha1 of imageuuid>.part file that is not growing, fir you do find one - then most likely you are seeing the same issue.

I posted about this on the mailing list a while back: http://www.gossamer-threads.com/lists/openstack/operators/39669
And also made a bug posting:  https://bugs.launchpad.net/glance/+bug/1340993

The bug posting included some scripts to duplicate/verify the issue.

In the end we moved ssl termination to haproxy on the same box as glance and the issues went away.
____________________________________________

Kris Lindgren
Senior Linux Systems Engineer
GoDaddy, LLC.


From: Alex Leonhardt <aleonhardt.py at gmail.com<mailto:aleonhardt.py at gmail.com>>
Date: Friday, January 9, 2015 at 2:42 AM
To: openstack-operators <openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>>
Subject: [Openstack-operators] Instances stuck in 'Build'

Hi,

recently I started having problems when creating instances - "every now and then" they seem to get stuck in a 'Build' state. After I terminate that instance and try again, it mostly goes through fine. This is happening across multiple projects/tenants and seems to affect all users.

Does anyone know this issue and/or has any suggestions or links or docs to read up on to troubleshoot this ?

Thanks!
Alex

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20150109/59a80c8d/attachment.html>