We run nova-metadata on all the compute nodes, then bind 169.254.169.254 to lo on each HV. This usually works with the standard iptables rule that nova-metadata add's. Worse case you just add it to the the default rules set for the compute node. Inside the images I think all you need to do is make sure that zeroconfig is turned off. So that the default route for 169.254.169.254 out eth0 is still there. I suppose you could also add a route via dhcp to always point out the 169.254.169254 out eth0. Worst case if the arp entry for 169.254.169.254 makes it out of the HV, you get automatic HA with all the other HV's on the same network who will respond to the ARP request for that vm. IE for us we typically have 43 other servers running metadata on the same network, so its a active, active, active, active config.
___________________________________________________________________ Kris Lindgren Senior Linux Systems Engineer GoDaddy From: Sam Morrison Date: Wednesday, September 30, 2015 at 7:24 PM To: Assaf Muller Cc: "[email protected]<mailto:[email protected]>" Subject: Re: [Openstack-operators] Neutron DHCP failover bug On 1 Oct 2015, at 10:52 am, Assaf Muller <[email protected]<mailto:[email protected]>> wrote: That's interesting. Looks like DHCP A/A only works if you use your (HA) routers to provide metadata, then. Yes that’s true, we’re not doing any L3 stuff in neutron yet. These are just shared external provider networks. Sam
_______________________________________________ OpenStack-operators mailing list [email protected] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
