NetApp Cheescake: failing to update connection information in cinder database when cinder failover-host is called

[chad morgenstern]

Following execution of cinder failover-host, I can only attach pre-existing volumes to instances if the nfs service is active on the prod vserver. If I disable the nfs service on the prod vserver, the cinder volumes list as attaching for a little while before timing out and going back to available state. With that said, volumes created after the failover can attach without issue.

db shows the fourth volumes as belonging to host 192.168.19.10 while the other are coming from .80. The fourth volume is the volume created after failover.