Sunday, June 2, 2013

Stopping HACMP cluster services without stopping applications

You can stop cluster services without stopping services and applications.
To stop cluster services without stopping your applications:
  1. Enter the fastpath smit cl_admin or smitty hacmp.
  2. Select System Management (C-SPOC) and press Enter.
  3. Select Manage HACMP Services > Stop Cluster Services and press Enter.
  4. Choose Unmanage resource groups.
No matter what type of resource group you have, if you stop cluster services on the node on which this group is active and do not stop the application that belongs to the resource group, HACMP puts the group into an UNMANAGED state and keeps the application running according to your request.
The resource group that contains the application remains in the UNMANAGED state (until you tell HACMP to start managing it again) and the application continues to run. While in this condition, HACMP and the RSCT services continue to run, providing services to ECM VGs that the application servers may be using.
You can tell HACMP to start managing it again either by restarting Cluster Services on the node, or by using SMIT to move the resource group to a node that is actively managing its resource groups. See Starting HACMP cluster services with manually managed resource groups for more information.
If you have instances of replicated resource groups using the Extended Distance capabilities of the HACMP/XD product, the UNMANAGED SECONDARY state is used for resource groups that were previously in the ONLINE SECONDARY state.
You can view the new states of the resource groups using the cluster utilities clstat and clRGinfo .
You cannot dynamically reconfigure (DARE) the cluster configuration while some cluster nodes have resource groups in the unmanaged state.
Warning about placing resource groups in an unmanaged state
When you stop cluster services on a node and place resource groups in an UNMANAGED state, HACMP stops managing the resources on that node. HACMP will not react to the individual resource failures, application failures, or even if the node crashes.
Because the resources of a system are not highly available when you place resource groups in an unmanaged state, HACMP prints a message periodically that the node has suspended managing the resources.
The ability to stop a node and place resource groups in an UNMANAGED state is intended for use during brief intervals for applying updates or for maintenance of the cluster hardware or software.
When you may want to stop HACMP cluster services without stopping applications
When you may want to stop HACMP cluster services without stopping applications
In general, HACMP cluster services are rarely the cause of problems in your configuration. However, you may still want to stop HACMP cluster services on one or more nodes, for example, while troubleshooting a problem or performing maintenance work on a node.
Also, you may want to stop HACMP cluster services from running without disrupting your application if you expect that your activities will interrupt or stop applications or services. During this period of time, you do not want HACMP to react to any planned application "failures" and cause a resource group to move to another node. Therefore, you may want to remove HACMP temporarily from the picture.