LPMC error

[radi_1]

Hi,
A customer has a very old server (G-class).It hanged and I had to reset it.The server booted without any error.
I checked the OLDsyslog.log file and found the following error:

vmunix:Detected low priority machine check
vmunix:error address 0x3237898
vmunix:LPMC type:SEDC(Ecc-corrected single-bit error).

It looks it is a RAM error.
I checked the RAM with sam & found that all system RAM capacity was read by kernel.The server is currently running with no problems.
Please advise.
Thanks

[Steven E. Protter · Accepted solution]

Shalom,

Low Proirity Machine check shows there was a problem. Usually only High Priority machine checks force a boot and crash.

I therefore suspect the system locked up due to memory starvation or corrupted server processes that were using that RAM chip.

I would run mstm/cstm or xstm and check that memory. This is likely to happen again unless you just had a very unlucky day.

Obvoiusly a system this old should be replaced with something you have a prayer of getting parts for if it breaks.

SEP

[VK2COT]

Hello,

A Machine Check is either a High Priority
Machine Check (HPMC) or a Low Priority
Machine Check (LPMC).

Software is executed to locate the source
of the failure and take appropriate recovery
action. In an LPMC, no system state is lost,
since the error was completely recovered and
is reported to software only for logging purposes.

You are OK for the moment. Since the
hardware is too old (G-class) and the
O/S too, I would recommend to start planning
the migration and upgrade.

End of Contract Availability for G-class
servers was on 29th of February 2004
(interesting! it was a leap year :)).

If no action is taken, you stand a good
chance to loose the whole server at some
time in the future. When and if it will
happen - it is difficult to predict.

Cheers,

VK2COT

[Bill Hassell]

LPMC type SEDC means that an error occurred in RAM but the hardware was able to automatically correct the problem. It is normal to have a few of these every few weeks. Unless you are running 10.20 with the online diagnostics installed, you won't be able to tell anything else about the error without running offline diagnostics (for a long time).

Do your customer a favor and recommend an upgrade, or at least make sure they have purchased 2 spare machines to use for parts (assuming this G-class is used for production work). It will very difficult to find replacement parts because this is so old.