Montecito Cache Errors

[Andrew Young_2]

Hi All

I have a number of rx7660's some with single and others with multiple cell boards as well as a number of rx2620's and rx4640's. All are multiple CPU servers.

We seem to be getting intermittent cache errors on the CPU's in all these units. There is not a high incidence of these just one a month spread on a random CPU.

I don't think its cause for concern, but I just need to confirm this. Of course EMS logs this as a MAJOR_WARNING, but the local HP techs seem to think EMS is being overzealous.

Of course the problem was noticed when an HP rep called at 1am on a Monday wanting to replace a CPU called in faulty by ISEE.

Regards

Andrew Y

[Sameer_Nirmal]

I am glad you did post another thread with details to get more insight of the issue.

I hope my reply in your another post http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1154618 would help.

[kris rombauts · Accepted solution]

The cache errors logged are not an issue, with the larger cache sizes in today's cpu's a higher (relative) number of cache errors are expected.
Those cache errors are so called CMC's (correctable errors) and will not cause an issue at the higher (OS) level since they are already corrected at the H/W level. If more then the allowed (per the CPU specifications) number of cache errors occur within a certain timeframe, the prefailure monitor (term used in the Windows OS management agents world, not sure of how this is called under Unix/EMS) will log a warning/error to let you know the CPU crossed the number of cache errors and it will advise you to replace the CPU pro-actively to prevent a OS issue in case the CPU would produce more errors over time that could influence the performance and stability of the system.

As long as the error threshold is not crossed there is nothing to worrie about.
From reading the thread(s) on this subject it looks like ISEE isn't filtering out those errors as it should.

[Andrew Young_2]

Hi All

I have been dealing with the local HP reps and it seems that the original ES threshold of 10 errors per 24 hours was set too low. Newer versions of EMS and ISEE fix this. We are in the process of updating EMS and ISEE on all these servers.

Thanks for your assistance.

Andrew Y