Using IBM Storage Scale Erasure Code Edition for data mirroring and replication

The secondary replica with synchronous mirroring by using GPFS replication can be set up by using IBM Storage Scale Erasure Code Edition. IBM Storage Scale Erasure Code Edition starts to support this feature from 5.0.5.2 release.

For more information on basic concept of synchronous mirroring with GPFS replication, see Data mirroring and replication topic in the IBM Storage Scale: Administration Guide.

In a configuration that uses GPFS replication, a single GPFS cluster is defined over three geographically separate sites. This GPFS cluster consists of two production sites and a tiebreaker site. Each production site has a set of IBM Storage Scale Erasure Code Edition storage nodes to create recovery groups.

In each IBM Storage Scale Erasure Code Edition recovery group, vdisks created are assigned to one disk failure group. The GPFS file systems that are created on these vdisks have two disk failure groups to hold file system data and metadata. Specifying file system replication factor of 2 for both data and metadata maintains two replicated file system blocks in each disk failure group. These replicated file system blocks provide a level of data redundancy that allows each site to continue to operate independently when the other site fails.

IBM Storage Scale Erasure Code Edition has two configurations that need to be adjusted for the mirroring and replication setting:

• nsdRAIDReadRGDescriptorTimeout: The default value is 300 seconds. It defines how long the recovery group tries to recover the root log group in each attempt.
• nsdRAIDMaxRecoveryRetries: The default value is 1000 times. It defines how many times the recovery group tries to recover before the vdisk failure is reported.

The suggested values for IBM Storage Scale Erasure Code Edition configured with mirroring and replication are:

• nsdRAIDReadRGDescriptorTimeout: 15 - 60
• nsdRAIDMaxRecoveryRetries: 3 - 5