Description of problem: Customer has run into an issue where the NVP plugin is causing deadlocks on the Quantum database. Here's the upstream issue: https://bugs.launchpad.net/neutron/+bug/1204277 Version-Release number of selected component (if applicable): openstack-quantum How reproducible: 100% Steps to Reproduce: 1. Set up RHOS 3 with Nicira plugin. 2. Launch some virtual machines and configure networking. Actual results: Quantum database queries get slow - like up to 1 min. Expected results: Quantum database queries should be sub-second.
Created attachment 791503 [details] Proposed Backport This backport has only been smoke-tested. It was modified a fair amount from the upstream backport.
We did some testing this afternoon. This patch introduces a bug where routers aren't deleted properly from the environment. Please do not use.
Ofer - could you please verify that the backporting did not cause any regressions, the scratch build to test is specify in comment #6.
(In reply to lpeer from comment #7) > Ofer - could you please verify that the backporting did not cause any > regressions, the scratch build to test is specify in comment #6. Done - I've install the following packages on clean rhel 6.5 and then installed Grizzly/2013-11-14.2 ran my auto-test, all passed successfully: http://download.devel.redhat.com/brewroot/work/tasks/6259/6556259/openstack-quantum-nicira-2013.1.4-rhbz1002258.el6ost.noarch.rpm http://download.devel.redhat.com/brewroot/work/tasks/6259/6556259/openstack-quantum-2013.1.4-rhbz1002258.el6ost.noarch.rpm http://download.devel.redhat.com/brewroot/work/tasks/6259/6556259/python-quantum-2013.1.4-rhbz1002258.el6ost.noarch.rpm
Michael, we backported the fix and QE tested that there are no regressions. We do not have access to setup with the required software (NVP/NSX plugin), who can verify the bug?
Adding >=POST to next async release
The patch with the backport and a scratch build was attached we did not get any feedback on the patch. If we want this fix we need someone to verify it.
The customer found a temporary workaround by disabling the metadata service and believes the issue will go away when they move to Havana, and thus they closed the ticket.
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days