[PVFS-users] PVFS error at client side

[Tajendra Vir Singh]

Sorry Rob,

I messed up. I have been working on this for last couple of days in order to
get it working. I am also facing problems using mpich-1.2.5-2 with pvfs
switch enabled. 

During this once I used the /pvfs/testfile.dat vs pvfs:/pvfs/testfile.dat,
and both of them did not work. They both came out without doing anything at
the /pvfs file system.

When I got it working with some combination of switches, I got the hangup
problem which I reported to the list as I thought it is not due to mpich but
it is due to pvfs.

TV Singh
5919A Math Sciences bldg,
ATS, UCLA, Los Angeles, CA 90095-1557
Ph: (310)825-5698  

-----Original Message-----
From: Rob Ross [mailto:rross at mcs.anl.gov] 
Sent: Wednesday, February 04, 2004 1:22 PM
To: Tajendra Vir Singh
Cc: pvfs-users at beowulf-underground.org
Subject: RE: [PVFS-users] PVFS error at client side

On Wed, 4 Feb 2004, Tajendra Vir Singh wrote:

> Hello Rob,
> 
> I am going to ask about applying the patch from our system administrator.
I
> think his answer will be no as he gave me the installed pvfs systems
around
> last week of January. I saw the patch you posted was on Jan 29th.
> 
> I already tried using pvfs:/pvfs/tesfile.dat in place of /pvfs/tesfile.dat
> and it did not help. 

This is interesting.  You should definitely *not* get ls hangs after 
running with the pvfs: if things are set up correctly.  That implies that 
maybe there is something wrong with /etc/pvfstab -- have other MPI-IO 
programs run correctly?

The reason that this is true is that a pvfs: MPI-IO run does not use the 
kernel interface, so there is nothing really that it would do to put the 
kernel interface in a weird state.  And it definitely sounds like that is 
what is going on.

Rob