View Only

Remote LPM Overview

By Pete Heyrman posted Wed June 17, 2020 01:04 PM

PowerVM Remote LPMOverview
Towards the end of 2007, when HMC support for Live Partition Mobility (LPM) was first released in V7R3.2.0, both source and target managed systems had to be managed by the same HMC.  Starting with the V7R3.4.0 release one year later, the source and target could be managed by two different HMCs, a feature called Remote LPM.  

V7R3.4.0 SP2 improved upon this by allowing the two different HMCs to be at different code levels.  Yet this is still unfortunately a little-known feature, one that can be very useful in situations where circumstances prevent both source and target server from being managed by the same HMC, or when moving to Power8 from Power7 for example.

The underlying HMC logic is the same as LPM where one HMC manages both source and target servers, with the notable exception being communication between the two HMCs to coordinate that logic.

Remote LPM uses SSH as the communication channel between source and target HMCs.  During the LPM, the source HMC will send several commands indicating target-side tasks to the target HMC.  The target HMC never initiates any communication to the source HMC.  Once the target HMC finishes each task, it will return result data back to the source HMC.

Note that LPM between HMC and IVM is not supported.

How to Prepare
Before starting a remote LPM operation, login to the source HMC to setup the communication channel to the target HMC:
mkauthkeys –u target_hmc_user_name --ip target_hmc_ip  --passwd target_hmcuser_password

After running mkauthkeys once, remote LPM will work from both HMC GUI and command line.
After setting up the communication channel, you can test it:
mkauthkeys –u target_hmc_user_name --ip target_hmc_ip –test

 If you want to setup remote LPM between two HMCs such that either can be a source HMC, you can run the mkauthkeys command on both HMCs, or on just one with the –g option:
mkauthkeys –u target_hmc_user_name --ip target_hmc_ip --passwd target_hmc_user_password –g

How to Initiate
Remote LPM can be started and aborted from a source HMC’s GUI, REST API or command line:
migrlpar –o m –m source_managed_system_name –t target_managed_system_name –p migrating_partition_name –u target_hmc_user_name --ip target_hmc_ip

 The –u and --ip parameters are only needed for migration (-o v), validation (-o v), and recovery (-o r). They are not needed to abort (-o s).

How to Recover from a failed remote LPM
Always initiate the recovery process from the source HMC.  Recovery should only be initiated from the target HMC if the source HMC is down or inaccessible.  Should you initiate recovery from the target HMC, you'll also need to run recovery from the source HMC once it is back up to ensure a complete cleanup.

Recommended Practices
  • Use HMC V8R8.2.0 or later, as many improvements to handle concurrent migrations were introduced in V8R8.2.0.  If you use V7 HMC, make sure it's at V7R7.9.0 with the latest Service Pack.
  • Before remote LPM, ensure SSH is working between the two HMCs (see preparation section above).  Log into the source HMC and try to run any HMC command on the target HMC via SSH.
  • After setting up the ssh communication channel, don't try to make any change in the target user’s .ssh directory.
  • Synchronize the time between the two HMCs.  Should you encounter an issue, debugging is easier if the HMC times were synchronized.

Contacting the PowerVM Team
Have questions for the PowerVM team or want to learn more?  Follow our discussion group on LinkedIn IBM PowerVM or IBM Community Discussions