topic Re: Testing Memory in Operating System - HP-UX

Testing Memory

brian bakke — Mon, 04 Dec 2000 22:39:30 GMT

good afternoon,
we are running HP-UX on our production servers and we are looking at testing the memory on the servers to see what type of alarms corrupt memory will generate - does anyone have any ideas on how to go about doing this? TIA

Re: Testing Memory

Alan Riggs — Mon, 04 Dec 2000 23:36:32 GMT

No. I know of no way to simulate a failure on a memory card. But memory cards should be monitored with EMS (Event monitoring service). the state of your event monitoring can be checked with:

/etc/opt/resmon/lbin/moncheck

Re: Testing Memory

brian bakke — Tue, 05 Dec 2000 00:25:57 GMT

Thanks - I haven't been able to find any info on testing the memory, so I figured it couldn't be done. Would there be any benefit from changing the memory parameters in the kernal? TIA

Re: Testing Memory

CHRIS_ANORUO — Tue, 05 Dec 2000 08:49:06 GMT

The kernel modification for memory will be of help if you understand what result will be. Read the attached guide for kernel parameter settings and their meaning.
http://docs.hp.com/hpux/onlinedocs/os/KCparams.OverviewAll.html

Re: Testing Memory

Carlos Fernandez Riera — Tue, 05 Dec 2000 08:54:46 GMT

I supose that a unrecoverable memory error will force a reboot of the box.

Memory errors are managed by hardware that can realocate blocks an deactivate bad areas.

In order to test no more memory allocation erros you can write a simple C that allocate lots of memory.

Hope This help.

Re: Testing Memory

Tommy Palo — Tue, 05 Dec 2000 09:01:53 GMT

Apart from memory check occuring at boot, if you have fastboot disabled, you can use the "Support Tools Manager" (stm or xstm) to test or exercise the memory.

Re: Testing Memory

paul courry — Tue, 05 Dec 2000 16:29:51 GMT

This is handled by the hardware. You have ECC SDRAM memory which will correct single bit errors. Once too many errors occur at the same location the memory location is marked as bad and is no longer used. I'm sorry that I don't know the exact verbiage used, but I suspect that either your CE can look it up or possibly you may find it on HP's web site. It is extremely rare to see a problem on a system that has been is use for a while. If it happens at all it happens when it is first installed, usually within the first 30 days.

Above all else, do not try and simulate a failure, you can fry a board that way. I'm sure that your friendly local memory salesman could supply you with a bad board if you are bound and determined to try this.

FYI, ODE (Offline Diagnostic Environment) has memory tests available in it if you wish to take the system down and perform tests.

Re: Testing Memory

brian bakke — Tue, 05 Dec 2000 18:22:57 GMT

Thank you all for the replys. Is there anything that you could recommend for a test procedure that would allow me to test the servers for a memory failure, or is this something that doesn't even make sense to test?