Converting a non-replicated file system to a replicated file system

You can convert an existing non-replicated file system to a replicated file system.

Before you begin

  • To perform this task, you must be a Db2® cluster services administrator.

About this task

This procedure is applicable to both single-site pureScale® cluster and a Geographically Dispersed pureScale Cluster (GDPC). With the latter, this is part of the steps in creating a GDPC cluster that involves creating a minimal non-GDPC cluster with non-replicated file system as the primary step. To illustrate the steps that are involved with figure 1 or 2, assume that the cluster has file systems created on storage 1 only. In other words, redundancy group 2 and tiebreaker are not defined. The cluster is operating with a single storage controller.
Figure 1. Multi-site IBM Spectrum Scale replication topology
Multi-site IBM Spectrum Scale replication topology


To enable IBM Spectrum Scale replication in the cluster, do the following steps:

  1. Group all disks currently in each file system into redundancy group 1, by using the following commands.
    db2cluster -cfs -enablereplication -filesystem <fs name>
    Repeat the command for each non-replicated file system. In figure 1, assume that all three file systems are non-replicated with disks that reside on storage 1 only. To convert the disks to replicated, run the following commands:
    db2cluster -cfs -enablereplication -filesystem db2sd_1
    db2cluster -cfs -enablereplication -filesystem db2datafs
    db2cluster -cfs -enablereplication -filesystem db2logfs
    At the completion of this step, all existing disks in each file system are assigned redundancy group ID 1. Any additional disks added to the file system after this step requires a redundancy group ID specified. Also no user data is replicated yet because secondary storage is not defined.
  2. Add disks to redundancy group 2 and the file system tiebreaker group to complete the disk setup.
    db2cluster -cfs -add -filesystem <fs name> -disk <disk1,...diskn> -rdncy_grp_id 2 -fstiebreaker <disk> -host <TB hostname>
    Accordingly, for example, in figure 1, run the following commands to add the disks to each file system :
    db2cluster -cfs -add -filesystem db2sd_1 -disk /dev/hdisk4 -rdncy_grp_id 2 -fstiebreaker /dev/hdisk7 -host TB
    db2cluster -cfs -add -filesystem db2datafs -disk /dev/hdisk5 -rdncy_grp_id 2 -fstiebreaker /dev/hdisk8 -host TB
    db2cluster -cfs -add -filesystem db2logfs -disk /dev/hdisk6 -rdncy_grp_id 2 -fstiebreaker /dev/hdisk9 -host TB
    At the completion of this step, the disk layout of the replicated file systems are completed. New data added to the file system after this step is replicated across redundancy groups 1 and 2 automatically. However, data that existed in the non-replicated file system before conversion is not replicated automatically. The replication and possibly rebalance of existing data needs to be triggered separately and individually. Since these two operations are I/O intensive and the duration depends on the actual size of data, it is recommended to run them at off-peak usage hours. For more information, see Replicating non-replicated data and Rebalancing a replicated file system.