Summary: RE: file size limit issue?

From: Stephen Reeves <SReeves_at_epnet.com>
Date: Thu Sep 27 2001 - 13:49:07 EDT
I received some very helpful suggestions.

The problem is that we have a program that runs on NT that builds an output
file via Samba running on Solaris to and NFS server (Auspex and Netapps)
that initially starts out fast then slows to a crawl.  Where it begins to
crawl has proven to be predictable.  It right above the 4 gig mark.

o We upgraded Samba to 2.2.1a which was suggested.  Ran a test with the
results being the same.
o It was suggested Auspex may be a problem.  Thankfully we do have a NetApps
box.  A test was done on the NetApps box but unfortunately the results were
the same.
o Someone suggested checking ulimit on the Solaris box.  We have unlimited
set.
o I tried doing a straight NT to Unix copy.  Although it didn't look quite
right while it copied, it did complete a 5.5 gig copy.  I was able to verify
the integrity of the file.  It looked good.
o Not only did we try 2 different NFS vendors hardware we tried multiple
file systems.  All the same results.
o No problem Sun to Auspex.  We build > 4 gig routinely.
o No problem building locally on NT.
o Used a file system with 80 gigs available.  Space not an issue.


Thanks to all who contributed.  It's still an open issue but all the
responses helped to knock some of the culprits off the list.  Also helped me
to look into things I didn't think off.  If we sort this out within the next
couple days I'll inform the group of what we find out.


> -----Original Message-----
> From:	Stephen Reeves 
> Sent:	Wednesday, September 26, 2001 9:10 AM
> To:	'sunmanagers@sunmanagers.org'
> Subject:	file size limit issue?
> 
> We are experiencing a problem that appears to be a file size limit.  We
> can't seem to nail down what component may be the culprit.  What happens
> is we have a process that runs on NT and outputs a large file (5 gigs) on
> our NFS server (auspex) via Samba (Version 2.0.7) running on Unix(Solaris
> 8).  Around 4 gigs the process slows down dramatically.  It has run up to
> 12 hours before we kill it.  As a test we had the same process run locally
> on NT and output locally and it finished in 1hr and 5 minutes.
> 
> We have run this process from Windows 2000 and 98.  The Unix server is
> always Solaris 8.  We have tried one rev back on Samba with the same
> results (2.0.6).  We run the same production process on a nightly basis
> with smaller files being output without visible problems.
> 
> Any ideas greatly appreciated.  Thanks in advance for any one willing to
> take time out to respond.  I will be happy to summarize.
> 
> Thanks,
> 
> Stephen
> 
Received on Thu Sep 27 18:49:07 2001

This archive was generated by hypermail 2.1.8 : Wed Mar 23 2016 - 16:32:31 EDT