Things to know about HA Manager Core Group in WebSphere Application Server

Body

A Core Group (HA Domain) is a component of the high availability manager function. It can contain stand-alone servers, cluster members, node agents, administrative agents and the deployment manager.

Core Group

• Logical grouping of application servers.
• A set of Application Server processes over which direct High Availability relationships may be established.
• Core Group information is persisted in a configuration file.
• WAS install creates the DefaultCoreGroup, which can never be deleted.
• Server creation/deletion adds/removes entries from the corresponding core group file.

Core Groups - Formation Rules

• There can be N number of core groups in a cell.
• A core group may contain zero or more WebSphere processes (JVM’s).
  • There is a recommended limitation of 100 servers in a core group. Please refer to my earlier HA Blog for more information.
• A WebSphere process (JVM's) must belong to one and only one core group.
• A core group cannot span cells.
  • All the members of a core group must belong to same cell.
  • A cell can contain more than one core group.
• A cluster cannot span core groups.

  • All of the cluster members must belong to the same core group.

  • A core group can contain more than one cluster.

  • When you move one cluster member from a coregroup, all remaining cluster members will also be moved automatically to the new coregroup.

Why Do I need More than one Core Group ?

• Typically, the single DefaultCoreGroup definition will suffice.
• In a large WAS topology additional core groups may be required.
  • Scalability  
  • Performance
  • Existence of Firewall between cell members
  • Geographically separated cell members

Note: You might have to create a bridge between the coregroups in order for the WLM to work properly. WLM uses bridge service to pass the routing information between the coregroup servers.

• Also a placeholder for HAManager disabled servers.
  • Its highly recommended to move all HA Manager disabled severs to a one special  coregroup. Though you disable the HAManager on few servers in a coregroup, the heart beat between the disabled and enabled servers will take place which can cause more overhead to the CPU. Also with some servers disabled HA view may constantly be unstable and break.

Moving server between Core Groups

• Stop the Application Server or nodeagent.
• Perform the movement to the desired core group.
• Save and sync the configuration changes to the running nodeagents.
• Perform a syncNode operation on the node the server was moved.
• Start the server.

Note: Deployment Manager can't be moved from the ISC/admin console. You need to use wsadmin to move DMgr.

            Starting with Version 7.0, it is not required to have one Admin Process (nodeagent or Dmgr) in the core group. It is a restriction in V6.x.

It is recommended to stop nodeagent when you move around the coregroup members. All cluster members should be present in the same coregroup. If you have EJB's deployed on the cluster , make sure you have the node nodeagent also part of the coregroup for the EJB WLM to work properly.