[ale] Linux NFSv2 & >9000 MTU

Denny Chambers bugfixer at bellsouth.net
Thu Nov 2 21:20:46 EST 2000


What is your read/write block size set to? I had to set our server and
clients to a 1024 read/write block size to gain stability. It hurt
performance but it was the most stable.

Denny Chambers

         __
        / /    __  _  _  _  _ __  __
       / /__  / / / \// //_// \ \/ /
      /____/ /_/ /_/\/ /___/  /_/\_\
      ...for IQs GREATER than 98...

> -----Original Message-----
> From: owner-ale at ale.org [mailto:owner-ale at ale.org]On Behalf Of Neil C.
> Bright
> Sent: Thursday, November 02, 2000 12:55 PM
> To: ale at ale.org
> Subject: [ale] Linux NFSv2 & >9000 MTU
>
>
> Anybody out there have experience with gigabit ethernet cards running with
> jumbo frames (>9000 MTU).  I seem to be having problems talking NFS (as a
> client) when I do such.
>
> Pings (even large ones) go through, and some NFS traffic does as
> well.  Within
> a short amount of time, it'll loose contact with the NFS server
> and processes
> waiting on that I/O will get into a non-blockable wait.
>
>
> stats:
>   Intel Pro/1000 gigabit ethernet
>   Cisco catalyst 6506
>   redhat's 2.2.16-3 kernel
>
> +============ E3 75 6D 3F D8 CF B5 C7 79 D3 44 11 DE 08 9C 85
> ============+
>  Neil Bright       ncb at cc.gatech.edu      IHPCL / CNS System Administrator
>  (404) 385-0448                                       College of Computing
>  http://www.cc.gatech.edu/projects/ihpcl   Georgia Institute of Technology
> --
> To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in
> message body.
>

--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.





More information about the Ale mailing list