non-tech Question on V-class memory chip reliability

[Steve Post]

2 or 3 times a year, I get a single or double-bit memory error on my V2500 or V2600 machine.

When this happens, I call hp, they replace the chip, then I'm quickly back in business.

But I find it usual that a HP9000 computer would have such poor reliability for their DIMM chips.

The question I have is: Do any other V-Class users have similar problems with their V-Class machines?

steve p.

[Mark Grant]

At my previous job we had two V class machines and they too were rather unreliable. We had to change memory a couple of times in the year and half I was there.

THe most annoying occasion was when I was called at 2.00am in the morning to be told the machine was down but when I logged in it was fine. This happened three times in a row until I thought to check the syslog and found the machine was re-booting every ten minutes or so. This, also was due to memory errors.

I rather liked the V class machines but they are not reliable.

[Jeff Schussele · Accepted solution]

Hi Steve,

We have 12 V-2600 systems here & I'll certainly backup you up i.e. we get more SBEs from them than any other type system - by far.
Of those 12 we've had 3 multi-bit panics/reboots roughly per year & dozens of SBEs.
When we escalated this with HP we discovered that there was a bug in the way diags (STM) decodes the address. The bug causes the address to be reported as always being in locked memory - kernel space to be exact - which means it cannot be deallocated until next reboot where it is correctly decoded & deallocated.
This is what causes the EMS alerts to be issued.
HP has released patches for this:

PHSS_29343 - s700_800 11.11 Support Tool Manager Mar 2003 Patch

PHSS_29344 - s700_800 11.11 Support Tool Manager Jun 2003 Patch

But we have not yet had the opportunity to apply them yet as we can't get the downtime.
But yes, I have to agree that Vs sure seem to have more memory errors than any other type system & would speculate that it's related to the V architecture (Vs were NOT designed by HP - rather they were acquired) & the massive BTU output they generate.

Rgds,
Jeff

[Ted Buis]

The V-class typically had more memory chips than other HP systems of it's generation due to larger memory configurations and certainly more than many of todays systems of similar total RAM that use higher density chips. The V-class used many low density module to spread RAM across many ports in the crossbar to gain memory bandwidth. What I am trying to say, is that the V-class systems typically had a high memory part count. If you have twice as many RAM chips, your probability of failure is twice as high. HP-UX 11i and I think 11.0 had the ability to handle some memory failures without a crash via memory page deallocation, but only if the STM is installed and set up. See http://www.software.hp.com/cgi-bin/swdepot_parser.cgi/cgi/displayProductInfo.pl?productNumber=B6191AAE

At the bottom of the link you can find another link to get STM for free.

[Alexander M. Ermes]

Hi there.
Using a V-class ( 2200 extended to 2500 )
since 1998.
Just one failure of memory carrier
and another failure of a RAM bar.
Rgds
Alexander M. Ermes