BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Virtualization administrators charged with keeping critical workloads running at all times can use vSphere Fault...
Tolerance, but there are several important requirements to consider when implementing this feature in a VMware environment.
First, administrators need to create and enable a vSphere high availability cluster of at least two suitable servers using the same ESXi and Fault Tolerance (FT) version. Each of the clustered servers must be able to access the same VM data stores and networks. FT logging and vMotion must also be enabled and configured.
When selecting host servers to include in a FT cluster, it is usually helpful to ensure that each server uses the same -- or very similar -- processors operating at roughly the same frequencies. A significant processor mismatch might cause performance problems as a slower system struggles to keep pace with a faster server. In addition, hardware virtualization -- such as Intel-VT -- must be enabled in each server's firmware. VMware FT may require additional licensing for each host in the cluster.
When implementing FT in vSphere, it is important to avoid unsupported or unique devices attached to particular systems. For example, if an original VM relies on a flash drive inserted into the server's USB port and an administrator attempts to shift execution to a duplicate VM on a server without the flash drive, a serious error may result. Make sure that all servers can access storage and the network equally. VM files must reside on shared storage, such as Fibre Channel SAN, iSCSI, NFS and NAS.
This sort of hardware discontinuity also extends to thin provisioning. All VMs require thick-provisioned virtual raw device mapping or virtual machine disk files of 2 TB or smaller. When disks are thin provisioned, a conversion to thick provisioning may be required, which demands powering off the VM.