An active-passive GPFS cluster

In an active-passive environment, two GPFS™ clusters are set up in two geographically distinct locations (the production and the recovery sites). These clusters are referred to as peer GPFS clusters.

A GPFS file system is defined over a set of disk volumes located at the production site and these disks are mirrored using storage replication to a secondary set of volumes located at the recovery site. During normal operation, only the nodes in the production GPFS cluster mount and access the GPFS file system at any given time, which is the primary difference between a configuration of this type and the active-active model.

In the event of a catastrophe in the production cluster, the storage replication target devices are made available to be used by the nodes in the recovery site.

The secondary replica is then mounted on nodes in the recovery cluster as a regular GPFS file system, thus allowing the processing of data to resume at the recovery site. At a latter point, after restoring the physical operation of the production site, we execute the failback procedure to resynchronize the content of the replicated volume pairs between the two clusters and re-enable access to the file system in the production environment.

The high-level organization of synchronous active-passive storage replication based GPFS cluster is shown in Figure 1.

Figure 1. A synchronous active-passive storage replication-based GPFS configuration without a tiebreaker site

This figure illustrates a synchronous, active-passive PPRC-based GPFS configuration without a tiebreaker site. The entire figure consists of two GPFS clusters, named the production site and the recovery site. They are attached using an IP network. The production cluster has five nodes: one node designated as the primary cluster configuration server, one node designated as the secondary cluster configuration server, two quorum nodes, and a non-quorum node. The production cluster has disks referred to as PPRC source volumes, and they are attached with shared NSD access. The recovery cluster has five nodes: one node designated as the primary cluster configuration server, one node designated as the secondary cluster configuration server, two quorum nodes, and one non-quorum node. The recovery cluster has disks referred to as PPRC target volumes, and they are attached with shared NSD access. The PPRC source and target disks are attached using the Peer-to-Peer Remote Copy function.