Setup/Installation:

This is not the official method to set up and test Disk Heart Beats but this is the method that I have settled in to while implementing them in the field. It provides a quick, recoverable method that does not impact the rest of the system. It does however require that you stop HA on both nodes. This has been used on systems with EMC Power Path devices. Proceed at your own risk.

Verify the required file sets are installed:
From AIX distribution media, bos.clvm.enh
From the HA distribution media, cluster.es.clvm.rte

1. Check with the IBM web site to ensure you have the latest code for these and rsct services.

2. Vary on all the volume groups to one of the nodes.
   'varyonvg <vgname>'

3. Convert the volume groups to Enhanced Concurrent Mode
   'chvg -C <vgname>'

4. Vary off the volume groups from the node.
   'varyoffvg <vgname>'

5. Repeat this on all nodes in the cluster. Now some may suggest that one could simply export/import the volume groups, this will also work but requires typing more characters.

6. Add the vg to a resource group and start cluster services

7. Use C-SPOC to add in the disk heartbeat network.

Problem Determination:
** one thing to remember, if using vpath devices you must use the "real" device.

1/ Enhanced Concurrent disks must be under the control of grpsvcs to work properly. Use the command 'lssrc -ls grpsvcs' to verify the number of disks currently under control of group services. The following example shows four ECM disks. There is a 's' group name and a 'd' group name for each disk.

Also check the status on all nodes that the disk is listed as "concurrent". If a disk is listed as "active" it is varied on in a non-concurrent fashion, and is incorrect.

If additional information is required you can look at the VGDA of each disk. Note the VG permission is set to passive-only.

If the disks do not look like this then it is possible that one on the nodes has the disk varied on in "active" mode. Look for Read/Write in the VG permission field. Normally only HACMP sets this condition.

2/ When the ECM disks initialize they must be brought online in passive mode:
"varyonvg -c -n -P vgname" should do the trick. Execute this on each node.
If there is a problem doing this, go back and vary off the VG on all nodes and retry the command. This should caus the cgoup services connection to be established for the disk.

3/ To activate the disk on a node use the command "varyonvg -c -n -A vgname" to cause the switch from passive to active. This is similar to the normal varyon of a volume group. When the disk is active the lsvg -o command should show it on line.

4/ To verify if the ECM disks are working the tool dhb_read can be used. It lives in /usr/sbin/rsct/bin/ directory. Group services must be running and the disk must be at least passive on both nodes for the dhb_read program to function.