[PVFS-users] What is the memory foot print of pvfs?

[Brent M. Clements]

Our compute nodes/client nodes seem to be falling off like crazy lately.
WE have users' running Gaussian jobs which are memory intensive and i/o
intensive. Well it seems that when their gaussian job hit's a memory limit
and exhausts all of the memory then does i/o using pvfs, pvfs craps out
and the entire system fails giving the following error:

Jun  6 20:10:02 n122 kernel: Unable to handle kernel paging request at
virtual address a0000000003bf000
Jun  6 20:10:02 n122 kernel: pvfsd[1986]: Oops 8804682956800
Jun  6 20:10:02 n122 kernel: --> memset [kernel] 0x240 <--
Jun  6 20:10:02 n122 kernel:
Jun  6 20:10:02 n122 kernel: Pid: 1986, comm:                pvfsd

What happens after this is that we can ping the machine but we can't
access it any longer. We have to essentially reboot the machine.

Anyone know what's going on?

The problem doesn't occur if they use the local /scratch disk space on
each compute node.


-Brent

Brent Clements


On Thu, 27 May 2004, Brent M. Clements wrote:

> How would I set the debug flag for the pfvs module? from modinfo, it's
> currently running at 0x0 which I believe means that debugging is off.
>
> Thanks,
> Brent
>
> Brent Clements
> Linux Technology Specialist
> Information Technology
> Rice University
>
> _______________________________________________
> PVFS-users mailing list
> PVFS-users at www.beowulf-underground.org
> http://www.beowulf-underground.org/mailman/listinfo/pvfs-users
>