Instant Capacity on Demand (iCOD) User's Guide for versions B.05.x

Special Considerations
Dynamic Processor Resilience (DPR)
Appendix B134
Dynamic Processor Resilience (DPR)
The LPMC monitor, within the Support Tools Manager (STM)
diagnostics, generates Information events for all cache errors that are
detected. After three errors (Threshold) have been detected on a
processor in 1440 minutes, or a 24-hour period of time (Period), the
monitor deactivates that particular processor, mark it for
deconfiguration on the next system reboot, and generate a SERIOUS
event. After the failed processor is deactivated, the LPMC monitor
attempts to activate one of the inactive iCOD processors, if any are
available. This method ensures the processing power of the system is
unchanged.
A default value of “three” is assigned to Threshold, except for the
PCX-W+ family of processors, which has a value of “five” assigned. The
default value assigned to a Period is 1440 minutes, or 24 hours, in all
possible processor configurations.
NOTE Dynamic processor resilience is supported only if the inactive iCOD
processors are configured for on-line activation.
V-Class Systems A V-Class system has a slight variation from: L-2000, N-Class,
Superdome, rp7410, and rp8400. This is because the V2500 and V2600
systems have dual processor boards. As a result, only the failing
processor is deactivated, and after the system reboots, the pair of
processors are deconfigured.
vPars In a vPar environment, processors deactivated by the LPMC monitor
must be manually replaced. See “LPMC Deactivations in vPars” on
page 95 for details.
Monarch
Processors
See “Failed Monarch Processors” on page 70 for details on replacing
failed monarch processors.