Juno controller random reboot or crash

reboot
crash

asked 2015-04-21 15:22:42 -0600

Girda
13 ●2 ●3 ●5

updated 2015-04-21 16:00:33 -0600

smaffulli
6981 ●38 ●68 ●102 http://maffulli.net/

Team, I am in the process of deploying a multi-node cluster with openstack juno on centos/RHEL 7.0. Finished the full deployment. Found 2 different issues. ( All the services are on controller & computes on each compute node )

I couldn't get the VM/instances to get internal IP assigned. Booting back on default kernel "3.10.0-123.el7" resolved the issue.
Once I resolve the above issue. Creating VM randomly crashes the controller node. Sometimes I can completely create a VM without any issues sometimes the first time I create it crashes the controller. Sometimes even without creating a VM it crashes the controllers.
- After the reboot, the controller node shows the compute-services down but the compute services are up and running on the compute nodes.

Tried with 3 different hardware sets, just to eliminate any hardware related issues. Don't see much of an error in neutron or nova related logs. Did anyone notice such issue?

Thank you Chakri

edit retag flag offensive close merge delete

2 answers

Sort by » oldest newest most voted

0

answered 2015-04-21 16:06:50 -0600

smaffulli
6981 ●38 ●68 ●102 http://maffulli.net/

I'd check on syslog for hints on the causes of the crash. Probably the crash is caused by a kernel module required by OpenStack Neutron or nova.

edit flag offensive delete link

Comments

The controller crashes. Just takes a nose dive so I don't see anything on the syslog. Thank you for the idea though.

Girda ( 2015-04-22 09:19:40 -0600 )edit

Is nova-compute running on the controller as well? If so try to disable the compute service on the controller.

jdexter ( 2015-04-24 11:19:17 -0600 )edit

0

answered 2015-04-24 11:21:53 -0600

flag of United States of America

495 ●3 ●12

updated 2015-04-24 11:23:56 -0600

To answer this part of your question

After the reboot, the controller node shows the compute-services down but the compute services are up and running on the compute nodes.

When rabbitmq is restarted it clears all of the queues, if the controller crashed and did not properly signal the old queues as being closed to the compute node, the compute service still believes that the old queues are active and continue to send to them. Restarting the services on the compute node will force them to check if their queue is available and if not they will create a new one.

This is due to the following bug https://bugs.launchpad.net/nova/+bug/...

edit flag offensive delete link

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-04-21 15:22:42 -0600

Seen: 253 times

Last updated: Apr 24 '15

Related questions