VMware High Availability (HA) monitors all virtualized servers and detects physical server and operating system failures. HA can improve the availability of the virtual SMU and make HNAS deployments more robust.
vSphere Fault Tolerance (FT) provides continuous availability for applications if a server fails.
The main HA and FT configuration options when used with virtual SMUs are as follows:
- vSphere vMotion® and
Storage vMotion®: Provide Manual and automatic migration of compute and storage without service interruption. The three vMotion scenarios are:
- Host-only migration (with shared storage)Moves VM execution from one host to another.
- Storage-only migration (single host access to two storage pools)Moves a VM’s disk image from one storage pool to another storage pool.
- Host and storage migrationA combination of both host and storage migration.
Risk of losing quorum: none to minimal. However, vMotion does not protect against an ESXi host loss.
- Cold standby SMU: In an ESXi HA cluster, if the ESXi host running the primary virtual SMU fails, a new instance of the primary virtual SMU starts on another ESXi host. The new instance uses the last updated disk image from the shared storage. Although recovery is fast, it requires starting the VM, which is not fast enough to prevent a quorum loss. If the HNAS cluster is healthy, an SMU HA failover does not affect its availability, but it does prevent access to
NAS Manager and the CLI while the new instance of the SMU starts.
Risk of losing quorum: high to certain.
- Hot standby SMU: With FT on, a secondary virtual SMU (on a different ESXi host) takes over immediately from a primary virtual SMU if the primary SMU fails. This requires a 10 Gbps FT logging network in addition to the normal network that connects the ESXi hosts and the
HNAS nodes. If the SMU serves as a quorum device, the failover should be within the five-second requirement before a quorum loss occurs. In this case, an SMU HA failover can occur without affecting the
HNAS cluster, even if one of the
HNAS nodes is down.
Risk of losing quorum: none to minimal.
In summary, HA provides a highly available virtual SMU, but failovers will cause a short-term loss of quorum. FT provides a highly available virtual SMU with a negligible chance of losing quorum.