Skip to main content

What an HA group is

Contributors

High availability groups use virtual IP addresses (VIPs) to provide active-backup access to Gateway Node or Admin Node services.

An HA group consists of one or more network interfaces on Admin Nodes and Gateway Nodes. When creating an HA group, you select network interfaces belonging to the Grid Network (eth0) or the Client Network (eth2). All interfaces in an HA group must be within the same network subnet.

An HA group maintains one or more virtual IP addresses that are added to the active interface in the group. If the active interface becomes unavailable, the virtual IP addresses are moved to another interface. This failover process generally takes only a few seconds and is fast enough that client applications should experience little impact and can rely on normal retry behaviors to continue operation.

The active interface in an HA group is designated as the Master. All other interfaces are designated as Backup. To view these designations, select Nodes > node > Overview.

HA Group Master Interface

When creating an HA group, you specify one interface to be the preferred Master. The preferred Master is the active interface unless a failure occurs that causes the VIP addresses to be reassigned to a Backup interface. When the failure is resolved, the VIP addresses are automatically moved back to the preferred Master.

Failover can be triggered for any of these reasons:

  • The node on which the interface is configured goes down.

  • The node on which the interface is configured loses connectivity to all other nodes for at least 2 minutes

  • The active interface goes down.

  • The Load Balancer service stops.

  • The High Availability service stops.

Note Failover might not be triggered by network failures external to the node that hosts the active interface. Similarly, failover is not triggered by the failure of the CLB service (deprecated) or services for the Grid Manager or the Tenant Manager.

If the HA group includes interfaces from more than two nodes, the active interface might move to any other node's interface during failover.