Overview
VAST Cluster supports the capability for NSFv3 clients to continue running applications against a mounted view without the need to remount the view when the primary cluster fails over to a replication peer. This capability is supported by synchronizing file handles between views on a replicated path on two or more replication peers.
To enable this feature, you enable global synchronization on views on the replicated paths. FSIDs and file handles then remain consistent across clusters in the event of a failover. You will also need to handle networking on the client side in the event of an actual failover so that the network layer represents the newly appointed cluster in a way that replaces the cluster that was active prior to the failover.
Configuring Globally Synchronized Views
In order to enable the seamless failover capability, the view that will be mounted by clients to access the replicated data on the source peer must have certain parameters synchronized with views on the replicated paths on the destination peers. This way, in the event that the source peer fails over to any of the destination peers in the replication group, client applications can continue working with their existing mount point.
To ensure that this capability is enabled on a given replicated path, you need to:
Create a view with global synchronization enabled on the replicated path on the source peer.
Create a view on the remote path on each destination peer, after replication between the source and destination is established and while the source peer is operational and accessible from the destination.
You are guided to create these views as needed in the configuration workflows for async replication.
When a view is created with global synchronization, the cluster checks for views on paths protected by the same replication group as the view path. The view's parameters are synchronized with any such views.
Performing a Seamless Failover on a Client
The mount point of an NFSv3 client directs traffic to one of the primary cluster's VIPs. Therefore, in the event of a failover to a different cluster, you need to take action on the client network to redirect traffic to one of the VIPs on the newly appointed cluster.
Note
Network handling on the client side is non-trivial, environment-specific and is outside of the scope of this documentation.