Erik Zandboer just posted a topic on the VMTN forum about an HA error he received when he updates his machines to 3.5 U2. The error was one almost everyone has probably seen by now “could not contact primary HA agent”. This is normally solved by pressing “reconfigure for ha” or disabling and enabling HA. This wasn’t the case this time. after some research Erik discovered that the host file entry for the ESX host did not match the DNS name, one of them started with a capital while other did not. This caused HA to fail, after changing the hostname/dns name and a reboot everything worked fine again.
I can imagine this happens because of the fact that VirtualCenter is actually performing as a DNS/Hosts file for HA. Inconsistent naming has always been, and probably will always be a problem. So before upgrading, check your hostname and /etc/hosts file!
Previously, enabling VMware High Availability required DNS resolution of all ESX Server hosts in a High Availability cluster. This was done using configuring DNS records or by adding all of the host names and IP addresses to the /etc/hosts file on each server.
Starting with the ESX Server 3.5 Update2 release, DNS resolution or /etc/hosts file entries are no longer required to configure High Availability. The host name and IP address information will now be provided by the managing VirtualCenter Server. the source