vDNS is not consistent with >2 controllers

[tom murray]

This is with 1.20/63

I have a cluster with 39 vRouters and 3 controllers. The cluster uses a HA configuration with a virtual IP floating between the controllers. I am using Virtual DNS configured for my IPAM.

in this environment, only some of the virtual machine IPs are getting resolved, others cannot be found.
I notice the DNS records are not being set consistently on each of the control nodes. See the attached screenshots.

I understand that this can happen because each compute node is peering with just 2 of the controllers, so some controllers do not get updates.

[tom murray (tmurray-a)]

I was told:

The workaround for this issue is to configure the DNS servers explicitly on all the compute nodes. The config file is /etc/contrail/contrail-vrouter-agent.conf. Please configure DNS servers under [DNS] section in the config file. Commented example is already available in the config file. This needs to be done for all the compute nodes. Please configure the same pair of DNS servers on all the compute nodes. After making this change please restart contrail-vrouter-agent on each compute node.

[Deepinder Setia (dsetia)]

Discovery fixed policy is not handling service expiry (going down and coming back up) correctly. This bug requires fix for https://bugs.launchpad.net/opencontrail/+bug/1401304

[Ashish Ranjan (aranjan-n)]

We need to implement syncing of vdns data