[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Strange nfs problem

Paul Crossman wrote:
I have a customer with a rather strange NFS problem that I'm just plain
stmped on and I'm hoping that someone might have some enlightening ideas.

Assume the following small network diagram:

OEM System (a) / \ / \ / \ / \ /------------\ Fedora RedHat Core 2 9 (b) (c)

The following NFS scenarios are currently working:

NFS Servers are running on B and C.

	File on C, read from A
	Write from A to either B or C
	Read and write between B and C.
	File on B, read from A is causing problems.	

The goal is to run software on A from an NFS mounted filesystem on B.

When that transfer is attempted the following has been seen
- The command line hangs and will unlock at the unexport of the NFS
- The data transfer always results in a file that is exactly 44KB in
Size (45056 bytes)
- Smaller files transfer just fine.
- Files larger than 44KB transfer file; only if they are larger than
some (unknown) upper limit do they block after 44KB. - Files are 120KB sometimes work, and sometimes block.

Does anyone know what might cause this?

44K sounds suspiciously like the circular buffer driver problem on some Intel cards. Make sure the drivers on ALL of your netcards are up to date.

You don't say what the OEM system is, so it's hard to say if the problem
is there.  If possible, you might try ensuring all of the NFS systems
can handle NFSV3 and if so, try using TCP mounts rather than the default

	(on A): mount -t nfs -o tcp B:/export /mountpoint
- Rick Stevens, Senior Systems Engineer     rstevens vitalstream com -
- VitalStream, Inc.                       http://www.vitalstream.com -
-                                                                    -
-  Jimmie crack corn and I don't care...what kind of lousy attitude  -
-                 is THAT to have, huh?   -- Dennis Miller           -

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]