LPMC and resulting CPU States

CPU Conditions as the Result

of a LPMC Error in a vPar Environment

Version 2.0 10/13/02

Introduction

This white paper provides an explanation of what happens to the state of a CPU in both a non Virtual

Partitions (vPars) environment and a vPar environment. It is meant to convey a basic understanding of

what happens to a CPU if LPMC errors are detected specifically in a vPars environment. Please see the

white paper on Dynamic Processor Deallocation And Dynamic Processor Resilience for the details

on the actual process of deallocation and deconfiguration.

Technology and Notes

• Possible CPU states: Assigned Bound, assigned Floating, and unassigned Floating (which can be

used in iCOD systems).

• The Monarch CPU in a vPar will always be a Bound CPU.

• The CPU with a logical ID of zero is always the monarch.

• With iCOD and vPars, the LPMC monitor does not automatically replace a failed CPU by

assigning a new Floating CPU to the vPar; this must be done manually.

• Deallocation and Deactivation are synonyms and refer to a CPU that is still recognized by an OS,

but is not actively scheduling processes; such a CPU may still handle I/O interrupts.

• Deconfigured refers to CPUs that are disabled by firmware when the system/nPar reboots; such

CPUs are not visible to the OS or vPar monitor.

Non-vPar Environment

The LPMC Monitor will make a call to deallocate and a call to mark-for-deconfiguration the CPU in

question, when the LPMC error threshold has been exceeded.

• For monarch CPUs, the mark for deconfiguration will occur, but not the deallocation.

• For non-monarch CPUs, the mark for deconfiguration will occur AND the deallocation will occur.

In either case, the CPU is NOT completely taken out of USE until the system/partition is rebooted.

vPar Environment

On a system with vPars, the behavior for the Bound Monarch CPU is the same as in a non-vPars

environment. Each vPar has only ONE Monarch CPU that is one of the Bound CPUs (i.e. a vPar can

have more than one Bound CPU). If the LPMC error threshold has been exceeded on the Bound

AH, Solution Architect, Enterprise Shared Services, Competitive Sales and Presales

PAGE 1
CPU Conditions as the Result of a LPMC Error in a vPar Environment 1 Version 2.0 10/13/02 Introduction This white paper provides an explanation of what happens to the state of a CPU in both a non Virtual Partitions (vPars) environment and a vPar environment. It is meant to convey a basic understanding of what happens to a CPU if LPMC errors are detected specifically in a vPars environment.
PAGE 2
CPU Conditions as the Result of a LPMC Error in a vPar Environment 2 Monarch CPU, the LPMC Monitor will make a call to mark-for-deconfiguration. The LPMC Monitor will not attempt to deallocate this Monarch CPU. Upon vPar OS reboot, the Bound Monarch CPU will reappear in the vPar as an active CPU (however it is still marked for deconfiguration). Upon vPar Monitor reboot (i.e. reboot of the hard partition), PDC will deconfigure the Bound Monarch CPU and the vPar Monitor will no longer see it.