HP9000/L2000 or N4000 DUAL PROCESSOR

[Katsu39]

Hello everyone,
There are two or more processors on the HP9000/L2000 or N4000 system, which mean dual or multi processor system.
I think the system will be rebooted if one of processor has HPMC or any other error.

1. After it is rebooted, the system will start without the processor which has error?

2. Which level error will be HPMC?

3. What happend to the system, if error occur on "Monarch Processor"?

4. What happend to the system which has 4 processors,if 3 of them are failed?

Regards;

Katsu

[Kevin Lamb_2]

Hi,

I beleive with the L-Class, if you loose a single CPU on a multi CPU board the server will fail, on reboot it should deallocate the failed CPU; however if this is the CPU in slot 0 (Monarch CPU) the server will fail to reboot.

If your server has 4 CPU's and 3 fail, as long as this is not the Monarch, the machine will reboot with out the other 3 however performace will be hit, and you would need HP to check why all three failed.

Again, the above is only what I understand to happen, luckily I have never had this problem.

Kev,

"I'd rather be flying"

[Cheryl Griffin]

On a multiprocessor system, a failed processor can register both LPMCs and HPMCs which will crash the system.

At the time of the crash, information regarding which processor caused the failure and details of what each processor was running at the time of the crash are saved in the dump.

[Tobias Hartlieb]

Hi,

generally, after an HPMC, all CPUs will boot up again - if possible! Only if the CPU has a permanent failure, the CPU won't come up anymore and will be deconfigured. This is independent of being the Monarch CPU or not.
Again, even if 3 out of 4 CPUs fail: the system will come up again with all 4 CPus unless it is a permanent error. Otherwise, it will start with only 1 CPU.

Regards.

Tobias

[Katsu39]

Hi All;
Thank you for replying me as quickly.
I had understood that the system would be re-started without failed cpu, after rebooted. (except it was monarch cpu)

The cause of panic are LPMC or HPMC.
The "cash parity error" can be LPMC or HPMC?

Thank you for your cooperation.

Regards;

Katsu

[Bill Hassell · Accepted solution]

This is quite hard to predict. There are thousands of potential failures that will cause an LPMC or HPMC. As far as I know, all HPMC's will crash the system and reboot. Whether a bad processor (or processors) will be automatically deallocated is very much dependent on the nature of the failure. Processor chips themselves have very low failure rates. What is more likely to fail are cache memory devices and memory controllers, memory and I/O busses and power supplies. The Monarch processor must always be available or the selftest will fail.

If the processors fail in such a way that they become deactivated, and the Monarch processor is OK, the system will boot and run as a single processor system. HP-UX doesn't know anything about the deactivated procssors so it uses what is available.

[Scot Bean]

I have an L2000 where once OnlineDiags detected an anomaly in one of 2 CPUs. This was not an [HL]PMC event.

The OnlineDiags automatically de-allocated the CPU and the machine continued operations.

[Katsu39]

Hell Scot;

Thank you for replying me.

This is very interesting story.

When you had had the OnlineDiags, which test or contents you had done?

And one of cpu had beeb de-alloceted by OnlineDiags, which mean the cpu has some problem, doesn't it?

Did you make sure some error log about the cpus which has been de-allocated, befor you had done the OnlineDiags?

Rgards;

Katsu

[Scot Bean]

The Online Diags routines run all the time if they are installed. (I am talking about EMS and STM). They have built-in threshholds for some repeating events.

Some CPU events exceeded a threshhold and the CPU was de-allocated. Machine did not reboot, 2nd CPU was still there and kept going with just 1 CPU.

[Katsu39]

Hello Scot;
Thank you for good infomation.

The CPU events which were occured on your system, weren't critical one at first?
Then after the events exceed a threshold,the CPU was de-allocated.
When it was de-allocated, how did you know that happen? For example event.log or syslog.log or console massage or e-mail to root something like a that?

What do you think or have an experience,If that CPU events were critical at once,the system would be rebooted?

Thanks
Katsu