Fail to launch the 10th instance on one compute node

asked 2014-04-08 09:33:05 -0600

Julian Zhu gravatar image

updated 2014-04-08 09:35:20 -0600

I have setup an openstack env (1 controller + 1 network node + 1 compute node) with packstack, and everything is ok for 9 instances running on my compute node. But when I try to launch the 10th instance, it hangs in the build state and neven moves on. I check the nova log and find nothing but the scheduler log below:

2014-04-08 22:16:32.017 3665 INFO nova.scheduler.filter_scheduler [req-8e2bdb2e-fac1-4904-82d0-d51f337fe108 d52f7f2cb8c54bd4b0ba8728f8f2208e 36fca0c3329a4ce68be6f1cb952d10e0] Attempting to build 1 instance(s) uuids: [u'd8fce202-e416-4002-9a55-5c1a05b9e07d']
2014-04-08 22:16:32.047 3665 INFO nova.scheduler.filter_scheduler [req-8e2bdb2e-fac1-4904-82d0-d51f337fe108 d52f7f2cb8c54bd4b0ba8728f8f2208e 36fca0c3329a4ce68be6f1cb952d10e0] Choosing host WeighedHost [host: nova-vhio, weight: 59277.0] for instance d8fce202-e416-4002-9a55-5c1a05b9e07d

I have already increased the instance number in project quota from 10 to 100, and there is still enough vcpus, memory and disk on the compute node for 10+ instance. I doubt it is a configuration issue but I can't figure it out. Can anyone help to look into this issue? Any suggestions are greatly appreciated.

edit retag flag offensive close merge delete

Comments

after I changed the quota in /etc/nova/nova.conf and restart nova-* service, the issue is resolved. It seems that the modification to project quota doesn't work at all.

Julian Zhu gravatar imageJulian Zhu ( 2014-04-08 22:16:06 -0600 )edit

What version of OpenStack was this issue seen on? The behavior seems unusual that tenant is able to spawn (quota - 1) instances in the first place! Please check the task_state of 10th instance when it is stuck in building state.

unmesh-gurjar gravatar imageunmesh-gurjar ( 2014-04-10 10:15:29 -0600 )edit