AIX LPAR (Capped/Uncapped) and LPM functionality

[matthias_van_b_1]

For several customers dealing with capped LPAR's (AIX) I am curious how FNMS is determining the uncapped/capped status as well as the assigned vCPU's . So we are looking at the outcomes of for instance an lparstat -i command and try to map the values to FNMS consumption figures. 

According to IBM it should be like this: "If the LPAR is Uncapped, it will use up to the number of Online Virtual CPUs. If it is Capped, it will use up to the Entitled Capacity or Virtual CPUs, whichever is lower". 

In addition to that, we are also searching for an explanation around LPM (live partition mobility) and possible effects on FNMS. Does FNMS take this into account and how exactly? 

I hope someone can elaborate on this topic. 

Thanks in advance!

BR, Matthias van Beeren

[tmullin]

Hi @matthias_van_b_1 ,

There's a property on the Hardware tab on an Inventory device called Partial number of processors. This value will always be used by the PVU licence reconcile as the metric for LPAR devices when it is populated. If you look in any NDI, the partial number of processors property will be set by the first populated property in this order -  VMCapacityInCores, VMCapacity, VMEntitlement and will be NULL if none of these are present.

If you want to look at the output from lparstat to work out how the partial processor count is populated. The agent will look at the Type and Mode properties, if Type is 'Shared' and Mode is not 'Capped', then it will take 'Online Virtual CPUs'. Otherwise it will take the 'Entitled Capacity'. If the partial number of processors is greater than the core count, then FNMS will use the core count instead as the partial processor count.

I don't have much experience with LPM but as far as I'm aware we do not support this at the moment. If anybody has experience with what happens when LPARs move host with LPM then maybe they can weigh in. I would recommend using the ideas portal to suggest this functionality.

Regards,

Tony

[matthias_van_b_1]

Thanks for your answer Tony! It helps and confirms our investigations so far. Hopefully someone will be able to address the LPM specific part.

Regards, Matthias