Post-upgrade task for cluster using IBM Spectrum Scale replication

During or after the pureScale instance upgrade from a version prior to V11.5, you are required to manually convert the tiebreaker redundancy group ID from 3 to 4. This task is an online operation.

Before you begin

  • Ensure that you have root access before you begin this task.

About this task

In a cluster with IBM® Spectrum Scale replication, the redundancy group IDs 1 and 2 are assigned to represent the two data replicas, and ID 4 is assigned to the tiebreaker. ID 3 is reserved for the third data replica in IBM Spectrum Scale.

The following procedure describes the steps to convert a tiebreaker redundancy group ID from 3 to 4 in a IBM Spectrum Scale replication.

Procedure

Verify if any tiebreaker disk in the file system is associated with 3 as its redundancy group ID using the following command run under root:

/usr/lpp/mmfs/bin/mmlsdisk <fs-name>
Note: The parameter must be prefixed by a /dev/xxx, the user can issue the command "/usr/lpp/mmfs/bin/mmlsfs all -A" to find <fs-name>

For example: /usr/lpp/mmfs/mmlsdisk /dev/db2fs1

If the redundancy group ID 3 is used for the tiebreaker, run the following command to convert it from 3 to 4:

  1. Create a stanza file with content:
    %nsd:
      nsd=<NsdName>
      failureGroup=4
  2. Apply the redundancy group ID change using the stanza file:
    /usr/lpp/mmfs/bin/mmchdisk <fs-name> change -F <stanza file>
  3. Rebalance the file system using db2cluster utility:
    db2cluster -cfs -rebalance -filesystem <fs-name>
    Note: Rebalancing of data can be I/O intensive and the duration depends on the size of data. It is recommended to run them at off-peak usage hours. The prefix `/dev/` must be removed from the parameter.

    For example: ./db2cluster -cfs -rebalance -filesystem db2fs1