Automate IO fencing keys re-registration of DMP devices inside LDOM after migration
Problem:
Need to manually run “vxdmpadm pgrrereg” command inside the LDOM with fencing configured in DMP mode after live migration .
Cause:
When a LDOM is configured with fencing in DMP mode and the LDOM is live migrated to another host, the initiator for the guest VM changes. The IO fencing keys that were registered from the old initiator needs to be re-registered from the new initiator.
The user needs to manually execute “vxdmpadm pgrrereg” command inside the LDOM that was migrated.
Solution:
Configure a parallel service group in the LDOM with an application resource. Use the scripts provided in this technote to configure the application resource. Copy the scripts locally to all nodes in the cluster. The scripts requires the virtinfo binary to be present inside the LDOM to detect if the LDOM has been migrated.
Example: The VCS cluster configured inside the LDOM consists of two nodes guestA and guestB. The scripts are copied to /pgrapp directory.
Log in to any of the cluster node
guestA# haconf –makerw
guestA# hagrp –add pgrsg
guestA# hagrp –modify pgrsg SystemList guestA 0 guestB 1
guestA# hagrp –modify pgrsg AutoStartList guestA guestB
guestA# hagrp –modify pgrsg Parallel 1
guestA# hares –add pgrapp Application pgrsg
guestA# hares –modify pgrapp StartProgram /pgrapp/pgrapp_start
guestA# hares –modify pgrapp MonitorProgram /pgrapp/pgrapp_monitor
guestA# hares –modify pgrapp StopProgram /pgrapp/pgrapp_stop
gusetA# hares –modify pgrapp Enabled 1
guestA# haconf –dump –makero
guestA# hasgrp –online pgrapp –any
When the LDOM is migrated from one host to another, the monitor script will detect the change in the host using the virtinfo command and execute the “vxdmpadm pgrrereg” command. If any errors are encountered it will be displayed in the VCS engine logs.