Thursday, May 21, 2009

VMware vSphere - Fault Tolerance

VMware High Availability provides protection against physical servers failures running ESX hypervisors. If one host in HA cluster fails then failed virtual machines are restarted on another alive host from cluster. It ensures the host has enough resources to fulfill requirements of newly booted virtual machines. It is able to monitor virtual machine activity by checking its heartbeat as well and in case of its failure to restart it.

The next logical step is fault tolerant virtual environment. VMware vSphere 4 can do it . It provides zero downtime and data integrity of virtual machines in case of physical server failure.

When you configure a virtual machine to be fault tolerant a secondary duplicate machine is created on a different host. Then, any operation performed on the primary machine is recorded and replayed on its duplicate. If the primary fails the secondary takes over and continues running without interruption. However, current version is not able to monitor applications running inside virtual machines but it should be available in future.

VMware Fault Tolerance or VMware FT as it is denoted is cool and must have feature but to implement it means to meet these requirements:
  1. VM (virtual machine) must be in HA cluster
  2. esx host ssl certificates checking has to be enabled
  3. VM has to be stored on shared storage
  4. VM's virtual disks have to be in thick format, thin is not supported already
  5. three VMkernel 1G ports are required, one for VMotion and two for FT
  6. FT doesn't support Virtual SMP, only single processor VMs are supported!
  7. physical hosts have to support hardware assisted virtualisation, no problem with recent servers
The most of requirements are common ones but points 4 and 6, for older servers 7 as well, are considerlable limitations. It's not so simple to implement VMware FT but I hope it will get better in next releases.

Finally, vSphere documentation is available at

No comments: