Description of problem:
Need to backport upstream Neutron fix 150110 "Improve DVR scale performance" to RHEL-OSP6.
Version-Release number of selected component (if applicable):
RHEL-OSP-6 (Juno) A4 (latest)
Steps to Reproduce:
1. Create 5 or more instances
2. Attach floating ips to the instances
3. Delete instances
The deletion causes ERROR in l3-agent (PortNotFound), which in turn triggers ERROR 'Failed to fetch router information' for the involved router. This in turn causes 'fullsync' of all Neutron routers, which is a very cpu intensive process in an environment with ~100-1000 routers. The high load averages (70-80% system time) cause the OpenStack environment to become sluggish as the Neutron L3 agent and neutron-server are running on the 3 HA controllers (co-located with other OpenStack services).
The deletion of multiple instances should NOT cause PortNotFound error, thus avoiding unnecessary full router sync.
The issue discovered/confirmed in IRC session with Assaf Muller.
According to amuller, the fix is contained in
Patch is present in OSP 7+, we need an OSP 6 downstream backport.
Checked with openstack-neutron-2014.2.3-19.el7ost.noarch
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.