•If the Cluster Manager is active on the local node, synchronization triggers a cluster-wide, dynamic reconfiguration event. In dynamic reconfiguration, the configuration data stored in the DCD is updated on each cluster node, and, in addition, the new ODM data replaces the ODM data stored in the ACD (Active Configuration Directory) on each cluster node. The cluster daemons are refreshed and the new configuration becomes the active configuration. In the HACMP for AIX log file, reconfig_resource_release, reconfig_resource_acquire, and reconfig_resource_complete events mark the progress of the dynamic reconfiguration.
•If the Cluster Manager is active on some cluster nodes but not on the local node, the synchronization is aborted.
8.5.3DARE Resource Migration Utility
The HACMP for AIX software provides a Dynamic Reconfiguration (DARE) Resource Migration utility that allows for improved cluster management by allowing a system administrator to alter the placement of resource groups (along with their resources—IP addresses, applications, and disks) to specific cluster nodes using the cldare command. The command lets you move the ownership of a series of resource groups to a specific node in that resource group’s node list, as long as the requested arrangement is not incompatible with the current resource group configuration. It also lets you disable resource groups, preventing them from being acquired during a failover or reintegration.
Dynamic resource group movement essentially lets a system administrator better use hardware resources within the cluster, forcing resource traffic onto one or more high-powered or better-connected nodes without having to shut down HACMP on the node from which the resource group is moved. Dynamic resource group movement also lets you perform selective maintenance without rebooting the cluster or disturbing operational nodes.
Using the DARE Resource Migration utility does not affect other resource groups that might currently be owned by that node. The node that currently owns the resource group will release it as it would during a “graceful shutdown with takeover”, and the node to which the resource group is being moved will acquire the resource group as it would during a node failover.
The following section covers the types and location keywords used in DARE resource migrations, and also how to use the cldare command and the -M flag to perform the migration.
Cluster Management and Administration 169
