[PVFS-users] FW: PVFS Hangups during concurrent read/writes

Rob Ross rross at mcs.anl.gov
Thu Aug 5 17:34:07 EDT 2004


What sort of network is this?  Are you getting errors on any of the 
interfaces?

Thanks,

Rob

On Thu, 5 Aug 2004, Brannen S Hough wrote:

> 
> 	Yet more information - not sure if this helps or not - but I set up
> a 2 node cluster.  I can run multiple test programs concurrently on the
> Manager node, I can run multiple test programs concurrently on the other
> node (that is only an IONode), but my test programs locked up when running
> one on each of the nodes - one doing reads and the other writes.
> 	The strange thing about that is they aren't actually hung.  After a
> long time (couple of hours) one of the two started up again, right where it
> left off, and ran at about normal speed to completion.  The other struggled
> slowly though a couple of more files, then stopped again for good, and was
> in the same place when I killed it a couple of hours later.  This gets
> weirder and weirder.
> 	Are people using PVFS more as a file server (where writes are
> seldom, and most accesses are reads)?  Any suggestions are welcome.
> 
> 	- Brannen
> 
> > -----Original Message-----
> > From: Rob Ross [mailto:rross at mcs.anl.gov]
> > Sent: Tuesday, August 03, 2004 2:58 PM
> > To: Brannen S Hough
> > Cc: pvfs-users at beowulf-underground.org
> > Subject: Re: [PVFS-users] FW: PVFS Hangups during concurrent read/writes
> > 
> > Hi Brannen,
> > 
> > Thanks for the problem report and for trying the newest prerelease before
> > getting back to us!
> > 
> > What exactly is your test program?
> > 
> > Does this only happen when you have the two iods running on the same
> > node?
> > 
> > It would be helpful to us for you to recompile with "-g", attach to the
> > pvfsd, and get a stack dump.  Also some debugging output could help... try
> > 0x077 and see if that gets much into the pvfsdlog file as a start.
> > 
> > The iods don't coordinate read and write operations in a way that would
> > cause deadlock, so that shouldn't be it.  The mgr just serializes
> > everything it does, so that should be ok too.  I'm not sure what is going
> > on quite yet...
> > 
> > Rob
> > 
> 
> 
> 
> 
> 


More information about the PVFS-users mailing list