PCI Parity Error 1

[mikes999]

We are runnuing company's Groupwise6 mail server on the NW5.1, SP3 Netserver LH 3r box. Several days ago it started messages like:

- PCI Parity Error on Bus/Device/Function 0058h
- PCI System Error on Bus/Device/Function 0058h

with corresponding critical errors from Netware. Appears that the error message generated almost immidiately after loading GWIA, although subsequent heavy usage of GWIA doesn't cause any additional messages (but sometimes does - up to 8 - 12 messages a day). The server doesn't crash, and mail appears to be working.

Neither HP, nor Novell are of any help.

Any suggestions are greatly appreciated

 

[yvfwc]

What else is running on the server? Have you restarted the server without loading other programs and found the error only starts after manually loading GWIA? Any other changes?

Cheers,

Mark ;-)

 

[TheLad]

This is definately a hardware error. I would start by getting some downtime on the server (or after an abend) and reseat both memory and processor(s).

If you can, also check in the system BIOS to see if you can discover the device/process running on 0058h as this may give you more clues.

This is not a NetWare error. -----------------------------------------------------
"It's true, its damn true!"
-----------------------------------------------------

 

[mikes999]

Mark - thanks.

Nothing much; mostly NDS related NLMs, SQL connector, and Inoculan - this is primarily mail server. The only change done to the box that (somehow) could be timewise correlated to the problem, was number of mail sockets change (server started responding extremely slow all of a sudden) from 5/3 to 8/5. And the errors did not start immidiately after that - the whole day or two later, so this correlation is also questionable.

You are correct in your assumption, that an error (usually two: Parity and System) pops up almost instantly after GWIA is manually started (our normal procedure); happens 9 time out of 10.
Any ideas?

 

[mikes999]

TheLad - thanks.

My initial thoughts exactly; I removed all devices from PCI slots but NIC, disabled all IRQ's but slot 3's (NIC's), and the only device that MIGHT use that address is a Latency Timer (no idea, what it is) in BIOS - moved it to 00C0h range. So far, no good.

Besides, a bunch of other "victims of the bug" think, that it is somehow related to a SP3 installation; HP even suggested to me to rollback to SP2 (which I will never do on a live, production and the only mailserver in the company).

Check out this thread:

http://forums.itrc.hp.com/cm/QuestionAnswer/0,,0xec55af48a9e5d5118ff40090279cd0f9,00.html

and thanks!

 

[yvfwc]

That is an interesting thread... Have you tried any of the fixes it mentioned? Have you thought about aplying SP4? It seems as though there are some combinations that work, it finding the right one. It seems to be hardware specific but caused by a mixture of drivers and/or apps. It's not very promising but I think I would try the solution on that thread by Ed Martens.

Good luck,

Mark ;-)

 

[mikes999]

mark,

yes, I've tried some (disabling IRQ's, for example). But all suggestions taken from there seem contradict each other. I will not apply any SP's (for the reasons stated above) until there is some hint of a solution; BTW, there seems to be another (unconfirmed) implication: web access to mailbox is krippling. After reboot, it's fine (it happened only once, and maybe unrelated).

As of your wishing me "a good luck" - same here, and, looking at the nastiness of the bug - we definitely need it!

A lot of it...

 

[mikes999]

Here is the update:

I am still getting an average 2 errors a day, no PCI messages but Netware critical errors (identical to the ones PCI related in the past); nothing in SYS$LOG.ERR, but a cutdown version of the same messages; everything works fine.

Is there any way to track the origin of a critical error in NetWare 5 ?

 

[TheLad]

Check SYS:SYSTEM\ABEND.LOG which will detail the affending process(es). -----------------------------------------------------
"It's true, its damn true!"
-----------------------------------------------------