Memory Fault

[Jim Adamson]

Can anyone help me with a memory fault on two of our HP-UX10.20 workstations.

When I try to use the top command on both hosts I receive the message 'memory fault (coredump)'. A core file is created in /dev/pty.

I have used swapinfo -m -t and only 3% of swap is being used. The vmstat -n output shows the fault /sy level at 30,000+ which is very high compared to machines that are working normally.

I have rebooted one of the workstations but I still receive the 'memory error' message for top.

This problem could only have started on these machines over the last couple of days. Any help with this will be greatly appreciated,
Thanks,
Jim.

[John Palmer]

Jim,

It's a common misapprehension that 'Memory fault (coredump)' refers to system memory.

What it usually means is that the program has attempted to execute an instruction that refers to an invalid memory address - in other words, a bug in the program.

If you run the command 'file core', this will tell you what type of error it was eg SIGSEGV. See 'man 5 signal' for some more information about the signal types.

Regards,
John

[John Palmer · Accepted solution]

Further to my earlier reply, you should ensure that you have patch PHCO_14306 for 10.20 installed as this does cure the symtoms that you describe.

Regards,
John

[Lasse Knudsen]

Hi there,

John is right about the cause of "memory error". In order to fix your problem apply the latest patch bundle. I am not sure exactly which patch fixes your problem. We have a very old workstation here that does the same with 'top'. It has a old (3 year) patch bundle and Y2K patches applied, so we are going to fix it using a patch bundle.

[melvyn burnard]

I believe you may want to install the top patch
PHCO_14306 obtainable frokm this site.