SvSAN has multiple deployment options for the witness to suit different requirements, these include:
Shared remote witness – This enables the minimum amount of IT infrastructure equipment that provides high availability (2 servers) to be located at the remote site. The witness is located at another central site (datacenter/HQ) and accessed over a WAN link. A single witness can be shared between hundreds of SvSAN clusters in remote locations.
Local witness – Similar to the previous configuration there are two servers each with an SvSAN VSA installed. This time however, the witness is hosted on a third physical server or virtual machine (outside the SvSAN HA cluster) located at the same site. This configuration is for environments which are totally isolated and have limited or no external network connectivity and need to protect against split-brain scenarios.
Multi-node SvSAN cluster – SvSAN can be deployed in multi-node clusters containing three or more servers. In this deployment scenario one of the SvSAN VSAs acts as a quorum for other cluster members. For example, with a 3-node cluster with VSAs A, B & C:
- When a mirror is created between VSAs A & B, VSA C acts as the quorum.
- When a mirror is created between VSAs A & C, VSA B acts as the quorum.
- When a mirror is created between VSAs B & C, VSA A acts as the quorum.
More information on multi-node SvSAN clusters is available in this white paper.
No witness – The final option is to deploy two servers at the remote site with no witness. In this configuration it is possible to enter a split-brain scenario in the event of loss of network connectivity between servers or if the servers are rebooted simultaneously. To reduce the chance of split-brain occurring, best practices should be followed. These include providing resilient network connections between servers, using quality components and using multiple, redundant power supplies.