NFS consumes network bandwidth

[Mike Eubanks]

On Sat, 2005-11-19 at 21:41 -0800, Mike Eubanks wrote:
> I'm really not sure where to start on this one.  No changes have been
> made on the server for some time.  Both are running 5.4-STABLE.
> 
> On the client (my workstation) I cleaned out my home directory to
> freshen up my Gnome desktop.  Shortly thereafter, the network load began
> to exhibit a consistent patern of stepping from 0% usage to roughly 80%
> usage, something it hasn't done in the past.  Every time I use an
> application such as Mozilla, Evolution, etc. (in Gnome 2.10) the network
> usage drops immediately to 0% for a short period, say 10-30 seconds, and
> begins to accumulate gradually again.  On the other hand, network load
> does not fall off when running any process as a daemon, or from the
> shell.
> 
> I used `tcpdump' to dump the packets being sent on the interface.  It
> appears NFS is generating all of the traffic, although, I'm not sure why
> because the actual mounts are not heavily used, or at least haven't
> demonstrated this behaviour in the past.  I unmounted all NFS file
> systems and the network load dropped off completely and did not return.
> 
> I did move the mount points directly to my home directory, rather than
> linking to a mount point that existed in /mnt.  This is about the only
> other thing I can think of, altough, I'm hoping someone could tell me a
> bit more before I reverse the changes.
> 

I reversed the mountpoint changes, although, it still exhibits the same
behaviour.  Below is an illustration of the current setup.

   server:share1  --> maps to -->  client:/mnt/myname/share1
   server:share2  -------------->  client:/mnt/myname/share2
   server:share3  -------------->  client:/mnt/myname/share2

The link from home directory on client machine looks like the following:

   /usr/home/myname/mnt/imports -> /mnt/myname

After I access the NFS filesystem, the load increases to a constant 80%
as described above.  Below is a snippet of the `tcpdump -c 1000 -vv -i
xl0' command for all traffic.  I removed the timestamps and abbreviated
the client and server host names.  The only two hosts talking are the
client and server, and all traffic is identical to the following.

client.220312819 > server.nfs: 96 fsstat [|nfs]
server.nfs > client.220312819: reply ok 168 fsstat POST: DIR 755 ids
1001/0 [|nfs]
client.220312820 > server.nfs: 96 fsstat [|nfs]
server.nfs > client.220312820: reply ok 168 fsstat POST: DIR 755 ids
1001/0 [|nfs]
client.220312821 > server.nfs: 96 fsstat [|nfs]
server.nfs > client.220312821: reply ok 168 fsstat POST: DIR 755 ids 0/0
[|nfs]
client.220312822 > server.nfs: 96 fsstat [|nfs]
server.nfs > client.220312822: reply ok 168 fsstat POST: DIR 755 ids 0/0
[|nfs]
client.220312823 > server.nfs: 96 fsstat [|nfs]
server.nfs > client.220312823: reply ok 168 fsstat POST: DIR 755 ids 0/0
[|nfs]

-- 
Mike Eubanks <mse_software at charter.net>