L2000 processor dead

[Dan DeHaan]

I have an l2000 server with 2 processors. The Event Monitoring Service found a problem with processor-0 and deactivated it. It had 3 parity errors in 1440 minutes. A reboot stabilized the system on the other processor, but the faulty one is still dead. I have no HP Support for this server and need some help to determine it's health and prognosis.

[Chris Vail]

A question for you: is the system protected by a quality UPS? And have you had the UPS checked recently. Funky power can mess up a CPU like you mention.
I had a IBM RS6000 had very wierd, unpredictable CPU errors, but they were only during working hours. After MONTHS of search, we finally had the AC power to the data center checked. The expert located a water fountain on a different floor of the building that was improperly grounded, and identified that the power panel in the data center was grounded to a PLASTIC water pipe. Every time somebody pressed the button for a drink of water, 20,000 volts went through the motherboard of my server. Surge suppressors and UPS's are only as good as their ground. So it might be good to look into the AC and grounding issues of your system.

If it isn't this, you have few choices. You can find some way to replace the faulty cpu, or just pretend the errors didn't happen and re-enable the silly thing. Replacing the CPU without a maintenance contract is pricey--you could probably replace the entire computer for the same or less money. Purchasing a used CPU is dicey--the odds of getting a good one are completely random. The last alternative is to muddle through on 1 CPU only.

This might be a good time to contact your
friendly neighborhood HP rep and see what kind of trade-in they can offer on a later/greater system. They're wheeling and dealing these days and so you might be surprised at how reasonable a new system with warranty might be for you. An RX2600 with 1 processor is easily as fast as an older L2000 with 2, and these are pretty cheap.

Chris

[Dan DeHaan]

Thanks Chris.

The server is in our server room which is on top of a battery backup system tied to an emergency generator.

How can I re-enable the CPU? It's not a production server - it was supposed to be by this morning, but this ended that project - so I don't care about getting more errors right now.

[Joshua Scott]

What parity errors are we talking about here, single-bit or multi-bit?

If I'm not mistaken, you can re-activate the CPU from the BCH (pre-boot) menu. If you go into the configuration menu, I believe that there is a processor command to activate/deactivate a Processor.

If you are experiencing multi-bit errors, then the CPU has failed, and is probably failing self-test. You can check why it was deactivated on boot when it lists the CPUs and their status.

Josh

[Dan DeHaan]

The error message didn't say what type of parity error - only that it had 3 in the Instruction or Data Memory Cache.

I switched the processors and it booted fine. So, with the help of a tech, I enabled the down processor and did an exercise test on it. Everything seems to be back to normal now. I'll let it idle for a couple weeks before I bring it up live.