Top
PRIMECLUSTERGlobal Disk Services Configuration and AdministrationGuide 4.5
FUJITSU Software

A.2.34 Changing Over from Single Nodes to a Cluster System

The following two methods are available for installing the cluster control facility on one or more single nodes where classes exist and changing over to a cluster system. Normally, the method "a" is used, but the method "b" can be used to skip backup and restore.

  1. Deleting classes and re-creating classes after changing over to a cluster system

    Back up volume data if necessary, and delete the root class and local classes. After installing and configuring the cluster control facility, re-create the classes and volumes, and restore volume data as needed.

  2. Changing over to a cluster system without deleting classes

    Change over from single nodes to a cluster system without deleting classes. However, you must understand and pay attention to the following points.


Important Point 1

A local class created on a single node cannot be used directly in a cluster system. When the cluster control facility is activated, the following error message is output to the system log and the GDS daemon log file, and the operation of the local class becomes unavailable.

ERROR: class: cannot operate in cluster environment, created when cluster control facility not ready

For details on resolution, see "(1) The error message "ERROR: class: cannot operate in cluster environment, ..." is output, and the operation cannot be conducted on the class class." in "F.1.9 Cluster System Related Error."


Important Point 2

Expanding the class scope after changing over from multiple single nodes to a cluster system, may output the following messages.

ERROR: class: class names must be unique within a domain

This error occurs when the name of a class created on a single node is the duplicate name of a class on another node. If this error occurs, rename either of the classes, and expand the class scope.

ERROR: class: volume minor numbers must be unique within a domain

This error occurs when the minor number of a volume created on a single node is the duplicate number of a volume on another node. If this error occurs, re-create either of the volumes, and expand the class scope.

The minor number of a volume can be viewed in the following manner.

# cd /dev/sfdsk/class/dsk
# ls -l
brw------- 1 root root 253, 33 May 6 09:00 volume1

Additionally, this error may occur when any lower level group exists in a class created on a single node or a class on another node. In this event, duplicate miner numbers cannot be checked with the method as shown above. Re-create all volumes and lower level groups in the class created on the single node and then expand the class scope.