Why variances in Utilisation in cores between CPU Pool and aggregated LPARs running in that Pool?

We have six uncapped IBM i LPARs, on a POWER6 595 server, and on 17-18 June we moved the LPARs from a 12-core Shared Processor Pool (SPP) to a 6-core SPP.  We use LPAR2RRD to see what the aggregated pre- and post- core utilisation is for all six LPARs, using the Server historical reports from 1 June through to the current date 23 June.  The sample rate is 1 minute (even though the data in the corresponding CSV files is 5-minute intervals), and we select 'CPU pool 1 : IBMi_12CPU', 'CPU pool 3 : IBMi_6CPU', 'LPARs aggregated', and the six individual IBM i LPARs. Then click Generate Report.
We then generate the CSV files for the two CPU Pools charts and the Aggregated LPARs chart.

There are two issues we see:
1. The Aggregated LPARs chart includes two extra LPARs that were not selected, in this case two VIOS LPARs (we only have a total of nine LPARs on this server).  Which is sort of OK because we can strip the data for those two LPARs from the CSV file we generate.  But why are the extra LPARs included when they were not selected?   (Hopefully not user error)
2. Why does the 5-min interval data for the CPU pools not match the corresponding intervals' aggregated data for the LPARs that run in those pools.  
Interval 20.06.17 18:40:00    
CPU pool IBMi_CPU6 has a utilisation in cores of 3.709.
For the seven LPARs running in the pool, the aggregated utilisation in cores is 4.944.  



  • And as soon as I posted this I realised I had made a mistake.
    Please close this comment for now and I will post a new comment once I have got my act together.

Sign In or Register to comment.