NFS Server

Rick Stevens rstevens at vitalstream.com
Fri Nov 11 20:40:32 UTC 2005


On Fri, 2005-11-11 at 15:31 -0500, Mike Poublon wrote:
> Rick Stevens wrote:
> 
> >On Fri, 2005-11-11 at 14:36 -0500, Mike Poublon wrote:
> >  
> >
> >>I'm working on setting up an NFS file server using Fedora Core 4. The NFS 
> >>shares will be accessed by compute nodes inside of a cluster computer. I've 
> >>successfully done this in the past, but due to some recent hardware changes 
> >>have run into a slight issue. The server (also head node of the cluster) has 
> >>Gigabit network cards. 9 of the compute nodes (NFS Clients) do not, they have 
> >>10/100 nics. There are also 5 compute nodes that do have gigabit cards. The 
> >>problem is this:
> >>
> >>When the nodes with gigabit cards mount an NFS share everything works fine. 
> >>When the 10/100 nodes try to mount, they are successful, but can't read large 
> >>files (~600mb). They display the error message
> >>
> >>server <hostname> not responding, still trying
> >>
> >>and sit there not realling doing much (as far as i can tell).
> >>
> >>The MTU is set at the default of 1500 on all of the NICS (gigabit and 10/100).
> >>
> >>If anyone has seen this problem and knows a way around it I'd appreciate some 
> >>hints.
> >>    
> >>
> >
> >It's not uncommon for autonegotiation to not really work between your
> >machines and the switch/router.  It would help if we knew what switch
> >you're using and the makes of the 10/100 cards.
> >
> I'm using Intel Pro 100 cards (drivermodule e100) and the switch is a 
> Netgear GS524.
> 
> >In lieu of that info, see if you can turn off autonegotiation on both
> >the switch ports connected to the 10/100 machines and the machines
> >themselves and force 100Mb, full duplex on both ends.  See if that
> >helps.
> >
> >  
> >
> I'd prefer not to switch everything back to 100Mb if I don't have to. 
> The reason this came up is because
> of a recent upgrade to the gigabit switch (it was a 10/100 switch 
> before). The thought was that having a gigabit
> path to the server would help out on performance.

I meant only force 100MB on the ports on the switch that are connected
to the 100MB machines.  However, I don't think you can do that on the
Netgear.  If you had an Extreme or Cisco, you could.  Still, see if
turning off autonegotiation and forcing 100MB FD on the machines buys
you anything:

	ethtool eth0 autoneg off speed 100 duplex full

----------------------------------------------------------------------
- Rick Stevens, Senior Systems Engineer     rstevens at vitalstream.com -
- VitalStream, Inc.                       http://www.vitalstream.com -
-                                                                    -
-                 IGNORE that man behind the keyboard!               -
-                                                - The Wizard of OS  -
----------------------------------------------------------------------




More information about the fedora-list mailing list