VMware vCenter Server Heartbeat

There probably isn’t much opportunity for people to play with vCenter Server Heartbeat given it’s steep entry price (~$12,000) is a bit out of range for most companies.  I recently had the opportunity to play with it while creating an environment to demo for a customer, so I thought I’d share.

vCenter Server Heartbeat is a application layer clustering technology used under license from Neverfail for your vCenter Server and database.  It also covers associated vCenter add-ons like Update Manager and the like. Previously, your availability options included placing vCenter in a VM and using HA to protect from physical host failure, or using Microsoft Cluster Services to protect from hardware, OS or application failures.  There are obviously downsides to either of these approaches.  The advantage of vCenter Heartbeat over these options is that it will cover the application and OS failure without the complication of Microsoft Cluster Services. It also adds performance-based coverage, which neither of the other options provides.

There are three deployment options: V2V, P2V or P2P.  The V2V option offers the advantage of being able to simply clone your existing vCenter server to create the secondary node.  With a P2V deployment you can essentially do the same thing with a P2V migration.  The P2P deployment requires a very similar hardware configuration for both nodes and a special cloning process using NTBackup.

You can configure the two nodes to be in the same LAN or in separate sites across a WAN.

In a LAN configuration, the recommended implementation is to create a heartbeat network to connect the two nodes.  Both nodes are identical copies of each other, so a packet filter is configured on the primary network connection to prevent the passive node from communicating on the network and causing IP or name conflicts.  When a failover occurs this packet filter is disabled on one host and enabled on the other.  The updates are sent between the nodes on the aforementioned heartbeat network to keep them in sync.

In a WAN configuration, if both nodes are on the same subnet then the above LAN configuration would apply.  If they are on separate subnets, then they will both be actively on the network with different names and IPs, and a split-brain avoidance module needs to be configured.

It can be a bit complicated to initially figure, but it definitely works.  Is it worth the cost?  That’s for you to figure out.  Good luck!