PLM (Partition Load Manager) is used to automatically manage CPU usage across a group of LPARs so that if one LPAR in a group is overworked, CPU can be taken from another less stressed LPAR in the same group and allocated to it automatically. The PLM servers for each frame are listed below:-

P595-3 - b38plm.brm.pri (10.84.48.145)
P595-2 - b22plm.brm.pri (10.84.48.175)
P595-1 - b17plm.brm.pri (10.87.48.130)

This document details instructions on how to setup and configure a PLM group for a group of LPARs..

The PLM software is installed via rsh, so it needs to be enabled on the host and connections allowed from the PLM server..

1. vi /etc/inetd.conf and uncomment the line below:-

#shell   stream  tcp6    nowait  root    /usr/sbin/rshd         rshd

Save and 'refresh –s inetd' to reread inetd.conf and enable changes

2. vi /.rhosts and add an entry for the PLM server as below

b38plm root 

Make sure that there is an entry in /etc/hosts for the PLM server..

3. smitty users and select Change / Show Characteristics of a User for the root user and change User can LOGIN REMOTELY(rsh,tn,rlogin)? to true

Now the client software needs to be installed, but first check that the server can connect with the client via rsh by running a remote command as root:-

rsh clientlpar-gbe date

If it returns the current date, then all is well, so then cd to /etc/plm/setup/ and install the client with:-

./plmsetup clientlpar-gbe root

After the software is installed, unset the options on the client lpar as detailed in step 1 to disable rsh access for root as it is a security risk.

The next step is to create the PLM group for the host (or add it to an existing group), this is defined by a file in /etc/plm/policies/ and named as the frame, for example on b17plm (in Laindon), the PLM file is /etc/plm/policies/P595-1.

vi /etc/plm/policies/P595-x

A PLM group entry is defined in the policy file as below:-

PLM_GRP:
      type = group
      cpu_type = shared
      cpu_maximum = 24
      mem_maximum = 0
      cpu_intervals = 3
      cpu_load_high = 0.9
      ec_per_vp_min = 0.3
      cpu_free_unused = no

cpu_maximum is the combined total of the CPU’s to be shared amongst the group, so if you have 5 hosts each with a desired CPU of 1, then your cpu_maximum is going to be 5.

When you create the group, just add an appropriate name (PLM_GRP in the above example) and set the cpu_maximum accordingly.

The entry for the clients of the PLM group will look like this:-

clientlpar-gbe:
      type = partition
      group = PLM_GRP
      cpu_shares = 128

Create an entry for each host like above and set the group according to the name you gave it.

Note – the CPU taken from each LPAR cannot fall below the minimum CPU value set for the LPAR profile, so if the desired CPU is the same as the minimum then nothing will happen because no CPU can be removed to give to another LPAR. Bear this in mind when setting up the PLM group as you may have to adjust minimum CPU values in the LPAR profile which will require a reboot.

Restart PLM with the following commands

/opt/support/PLM/stoplm.ksh P595-x
/opt/support/PLM/startplm.ksh P595-x

Check status of PLM with 'xlpstat -p /etc/plm/policies/P595-x 1 10' where x is the frame number, 1 is the interval and 10 is the amount of checks.

You can also tail the logfile for info on what PLM is doing which is in /var/opt/plm/. The logfile is named P595-x.log.Mth where x is the frame number and Mth is the current month..