Open Stack

We also run DHCP as active-active on both nodes and don’t do any failover at all.  Worst case, both DHCP agents respond to a client, but the lease info is the same from both, anyway.


From: Clayton O'Neill
Date: Wednesday, September 30, 2015 at 5:48 AM
To: Sam Morrison
Cc: OpenStack Operators
Subject: Re: [Openstack-operators] Neutron DHCP failover bug

We've seen similar issues with three network nodes, 2 dhcp agents per and automatic failover.  We were seeing spurious failovers on a regular basis and we'd end up finding out about it because an instance would fail to get a lease.  When we investigated, usually see the issue you describe.  Automatic failover is on by default in Kilo, so it makes this bug much worse.  We ended up turning off automatic failover because of this issue.  There are other fixes for failover in Liberty that were never back ported to Kilo.

On Tue, Sep 29, 2015 at 9:58 PM, Sam Morrison <sorrison at gmail.com<mailto:sorrison at gmail.com>> wrote:
Hi All,

We are running Kilo and have come across this bug https://bugs.launchpad.net/neutron/+bug/1410067

Pretty easy to replicate, have 2 network nodes, shutdown 1 of them and DHCP etc. moves over to the new host fine. Except doing a port-show on the DHCP port shows it still on the old host and in state BUILD.
Everything works but the DB is in the wrong state.

Just wondering if anyone else sees this and if so if they know the associated fix in Liberty that addresses this.

Cheers,
Sam


_______________________________________________
OpenStack-operators mailing list
OpenStack-operators at lists.openstack.org<mailto:OpenStack-operators at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20150930/7d8edc15/attachment.html>