EMS active fail over package

[Rene Mendez_4]

Hi

I configure MC/ServiceGuard cluster.
A package have (sybase, cobis) only startup and shutdown scripts "no monitor these"

The question is:
When exist hardware problem:
Memory
CPU
Celda
Fibre Channel

This events active the cluster ?

EMS active the cluster ?

Regards
Rene

[Patrick Wallek]

It depends on what type of problems you have.

If this is a multi-CPU system, then it is possible for a CPU to fail and the system to remain running. It is also possible for a CPU to fail and crash the system. If this is a single-cpu system and the cpu fails, then you are out of luck as the machine will crash.

In the case of RAM, it depends. You can have LPMC (low priority machine check) errors, which are typically RAM, but should not cause any problems. You can have HPMC errors (high priority machine checks) that will cause your system to crash.

Fibre channel errors should not cause you any problems accessing your storage. You MUST have multiple paths to access your disks. If one of the FC cards, or a FC cable, or a FC director or switch fail, then you should still have access to your disk. If you don't then the system isn't designed correctly.

There really is no way for MC/SG to monitor these. Even if it did, as most of these errors SHOULD NOT cause your machine to crash, thus the MC/SG package should keep running.

MC/SG is all about eliminating SPOF's (single points of failure) and multiple CPUs, multiple fibre cards and paths to disk, multiple power sources, etc. are all related and a loss of any one should not effect your machine.