Browse Prior Art Database

Method to Hot Spare failed processors in DLPAR or CUoD system

IP.com Disclosure Number: IPCOM000011974D
Original Publication Date: 2003-Mar-27
Included in the Prior Art Database: 2003-Mar-27
Document File: 2 page(s) / 45K

Publishing Venue

IBM

Abstract

Method to Hot Spare failed processors in Dynamic Logical Partitioned (DLPAR) or Capacity Upgrade on Demand (CUoD) system.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 2

Method to Hot Spare failed processors in DLPAR or CUoD system

In the current implementation of the pSeries GigaProcessor a fatal error in the Translation Lookaside Buffer (TLB), the Sector Lookaside Buffer ( SLB) , or Data-Effective to Real Address Translation Cache (DERAT) cause the processor to be unusable in a logically addressed system. These errors cause the logical to physical addressing mechanisms to be unusable and therefore the processor cannot be used in other than real address mode. The processor can function as long as address translation is not necessary, e.g. in hypervisor mode, however the processor is unusable to the current version of the operating system because it cannot run without address translation enabled,. Because it cannot recover in real mode the operating system and all applications must be terminated without any warning and consequently have no opportunity to gracefully terminate.

To help prevent such operations under these error conditions recovery methods may be implemented, however these methods have dependencies on the following: The number of processors allocated to the Logical Partition (LPAR).

The number of processors in a free pool, i.e. processors unassigned to a LPAR and

available for assignment.

These dependencies limit the recover as follows:

If the system has unassigned processors the automatic removal and replacement of a processor is possible without loss of function or performance. If a system has no unassigned processors but has more than a single processor per partition it is possible to recover without loss of function but only incur a loss of performance in the partition. If the termination of the OS in cases where the partition has only one processor.

The assumption is made that the...