<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7232.36">
<TITLE>RE: [MPICH] tricks for avioding locking problems</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2>OK, I understand what you are saying.<BR>
<BR>
Which hint disables the data sieving operation?<BR>
Is there a good online resource for hint documentation?<BR>
<BR>
Thanks for the help!<BR>
-Jim<BR>
<BR>
-----Original Message-----<BR>
From: Rajeev Thakur [<A HREF="mailto:thakur@mcs.anl.gov">mailto:thakur@mcs.anl.gov</A>]<BR>
Sent: Tue 6/21/2005 3:51 PM<BR>
To: Foucar, James G; mpich-discuss@mcs.anl.gov<BR>
Subject: RE: [MPICH] tricks for avioding locking problems<BR>
<BR>
Which file system you use matters because different file systems have<BR>
different consistency semantics, and we accordingly have to compensate for<BR>
it in order to provide complete MPI-IO functionality. If you are using NFS,<BR>
you pretty much need locks to work unless you are only doing reads from the<BR>
file. On a true POSIX file system, you need locks to implement shared file<BR>
pointer, the atomic mode, and the data sieving optimization for<BR>
noncontiguous writes (which can be disabled with an info hint). We have come<BR>
up with a way to avoid locks on a POSIX file system by implementing the<BR>
equivalent of locks with MPI one-sided communication, and have 3 papers on<BR>
it, but it is not in the ROMIO release yet. (Locks will still be needed on<BR>
NFS to disable client-side caching.)<BR>
<BR>
Rajeev<BR>
<BR>
<BR>
<BR>
_____ <BR>
<BR>
From: owner-mpich-discuss@mcs.anl.gov<BR>
[<A HREF="mailto:owner-mpich-discuss@mcs.anl.gov">mailto:owner-mpich-discuss@mcs.anl.gov</A>] On Behalf Of Foucar, James G<BR>
Sent: Tuesday, June 21, 2005 4:34 PM<BR>
To: Rajeev Thakur; mpich-discuss@mcs.anl.gov<BR>
Subject: RE: [MPICH] tricks for avioding locking problems<BR>
<BR>
<BR>
<BR>
Let's not focus on NFS at the moment. What I really need is a way, trick, or<BR>
hack that allows me to circumvent or disable file locking in general. Does<BR>
such a thing exist?<BR>
<BR>
Thanks,<BR>
Jim<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: Rajeev Thakur [<A HREF="mailto:thakur@mcs.anl.gov">mailto:thakur@mcs.anl.gov</A>]<BR>
Sent: Tue 6/21/2005 3:25 PM<BR>
To: Foucar, James G; mpich-discuss@mcs.anl.gov<BR>
Subject: RE: [MPICH] tricks for avioding locking problems<BR>
<BR>
One way to avoid this problem is to not use NFS :-).<BR>
<BR>
Do you have the lockd daemon running on all nodes?<BR>
<BR>
Rajeev<BR>
<BR>
<BR>
_____<BR>
<BR>
From: owner-mpich-discuss@mcs.anl.gov<BR>
[<A HREF="mailto:owner-mpich-discuss@mcs.anl.gov">mailto:owner-mpich-discuss@mcs.anl.gov</A>] On Behalf Of Foucar, James G<BR>
Sent: Tuesday, June 21, 2005 4:12 PM<BR>
To: mpich-discuss@mcs.anl.gov<BR>
Subject: [MPICH] tricks for avioding locking problems<BR>
<BR>
<BR>
<BR>
Hello,<BR>
<BR>
I was wondering if any of you know of ways to avoid the following error:<BR>
<BR>
> File locking failed in ADIOI_Set_lock. If the file system is NFS, you<BR>
> need to use NFS version 3 and mount the directory with the 'noac' option<BR>
> (no attribute caching).<BR>
> [0] MPI Abort by user Aborting program !<BR>
> [0] Aborting program!<BR>
> p0_928: p4_error: : 1<BR>
> exit 1<BR>
<BR>
This error, and similar errors, have popped up randomly on all many file<BR>
systems, regardless of whether they were mounted with attribute caching.<BR>
<BR>
I have tried using the MPI_MODE_UNIQUE_OPEN flag when opening files and some<BR>
other tricks, but nothing seems to get rid of this problem permanently on<BR>
all the various platforms that I need to support.<BR>
<BR>
What I really need is a general way to tell MPICH not to use any file<BR>
locking or at least some way to avoid this error.<BR>
<BR>
Thanks!<BR>
-Jim<BR>
<BR>
<BR>
<BR>
<BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>