Nova network is crashing hard [closed]

asked 2013-09-23 04:37:53 -0500

ArKam gravatar image

updated 2014-01-22 15:12:39 -0500

Evgeny gravatar image

Good morning everyone!

Following my white rabbit on the wonderfull world of OS, I just came with a really strange issue on my Nova Compute/Network node.

As soon as I create an Instance, the creation fail because of a Nova Network error.

My node is compose of the following:

2xXeon E5520. 32Gb Ram. 67Gb Local HDD. 1Tb iSCSI Equallogic node (Cinder volume). Fedora 19 Schrodinger's Cat. Openstack Grizzly release.

The Fedora 19 is a fresh minimal install, and OpenStack component are coming from the default Fedora Repository. All installation are proceeds using yum packet manager and then all configuration steps are those explain on the official OpenStack Grizzly installation manual for YUM Distribution.

I even double check the procedure with the RDO Installation manual and the RedHat installation manual.

I have installed the Qpidd C++ AMQP Broker, which is working great as cinder is available to fast and quickly create volume from Horizon Dashboard and/or from the local CLI.

Every nova module seems to working great, except the nova-volume which crash with the following stack on logs:

Nova Network Crash

Yeah, this is a big bad stack :D

Here is my Nova.conf file:

nova.conf

CLUNODE01 is the Cloud Controller which host Horizon/Glance/MySQL CLUNODE02 is the Compute Node which host Nova-Compute/Volume/Scheduler/Network/VNCConsole/etc

My local lan is on a 192.168.0.0/24 network, I've a bridge br100 link to my em1 network card and it address is 10.2.0.1/24.

My network on OS is the 10.2.0.0/24 network where 10.2.0.1 (br100) is the default gateway.

From my own search, it seems to be a bug, but all that I found is a bug coming from the Havana release that I do not use and a man which have the exact same error with the exact same image used.

Man with same error

Bug in havana

Explanation of the bug

If anyone have a clue, I'm on it :D

edit retag flag offensive reopen merge delete

Closed for the following reason question is not relevant or outdated by larsks
close date 2014-03-04 19:36:32.650281

Comments

I did test with CirrusOS and the exact same behavior happen... if it can help, I've had my Nova.conf on the question.

ArKam gravatar imageArKam ( 2013-09-23 05:30:37 -0500 )edit

Not too sure here but Looking at the bug, it could be an error with SQLAlchemy 0.8.x. Is it possible to downgrade SQLAlchemy to a suitable version(0.7.x) and test it?? Not entirely sure if this is the best solution, but it's worth a shot? :)

Sirushti Murugesan gravatar imageSirushti Murugesan ( 2013-09-23 11:06:17 -0500 )edit
1

Yep, that what I though, I'll do the try, as you said it, It worth a shot, plus, I'm on a lab test environnement ;-)

ArKam gravatar imageArKam ( 2013-09-24 02:19:25 -0500 )edit

Ha! Awesome :)

Sirushti Murugesan gravatar imageSirushti Murugesan ( 2013-09-24 02:24:09 -0500 )edit

Bad luck, I can't do that, if I try to do so, the following packages are dependant of this python-sqlalchemy version and override the older one as soon as I reinstall them :(

ArKam gravatar imageArKam ( 2013-09-24 02:47:57 -0500 )edit