HP Instant Capacity (iCAP) manual Replacement of Failed Cores

Replacement of Failed Cores

HP-UX LPMC and HPMC

If an active core fails with a Low Priority Machine Check (LPMC) in a partition with Instant Capacity, its processing capacity is replaced instantly by an inactive core, if any are available in the partition. The failed core is marked for deconfiguration during the next system reboot.

For additional considerations in a virtual partition environment, see “LPMC Deactivations in Virtual Partitions” (page 73).

If an active core fails with a High Priority Machine Check (HPMC), then upon reboot the failed core is deconfigured and its processing capacity is instantly replaced by an inactive core, if any are available in the partition.

NOTE: In both of the preceding scenarios, replace the failed core in a timely manner using your normal hardware support process.

LPMC Deactivations in Virtual Partitions

In a vPar environment, if the LPMC monitor deactivates a core, it automatically replaces the failing core with an Instant Capacity core from the free pool, if such a core is available. The failing core remains in the virtual partition until either the virtual partition or the virtual partition monitor is rebooted.

For more information about LPMC in vPars, see the various white papers at the HP Documentation web site (search for “LPMC”):

http://docs.hp.com

Failed Monarch Processors (HP-UX)

Monarch processors that are failing with an LPMC are not instantly replaced. When a monarch processor experiences an LPMC, the LPMC monitor marks the processor for deconfiguration; however, the LPMC monitor cannot deactivate the processor until the system is rebooted. Deactivation of a monarch processor is not possible because it is the controlling processor of the operating system (CPU 0). Therefore, the system cannot replace a (failing) monarch processor.

If your system has only one active processor, that processor is considered a monarch processor and it cannot be replaced on line. A reboot of the system is required to replace the failing monarch processor.

If there are multiple active processors in your system, one of them is designated as the monarch processor, and the other (nonmonarch) processors can be replaced on line. If the monarch processor fails, it cannot be replaced without a reboot.

Replacement of Failed Cores on OpenVMS

If a core is experiencing correctable errors, shut it down and start up another Instant Capacity core, thereby keeping the active-core count constant.

If a core experiences a fatal problem leading to a system crash, upon reboot you can start another Instant Capacity core, thereby replacing the failed core and keeping the active-core count constant.

Failed OpenVMS Primary Processors

An OpenVMS primary processor that is failing cannot be instantly replaced.

If your system has only one active processor, it is considered a primary processor and it cannot be replaced on line. A reboot of the system is required to replace the failing primary processor.

Replacement of Failed Cores 73