 | | | jumbo frames issue (9000 bytes MTU) | jumbo frames issue (9000 bytes MTU) 2005-07-29 - By Tom Sightler
Back
On Fri, 2005-07-29 at 12:39 +0200, wolf2k5 wrote:
> On 7/28/05, Lee Whatley, Contractor <lwhatley.ctr@(protected)> wrote:
> > We have some Sun nfs servers that use jumbo frames, and to get RHEL to
> > nfs-mount those filesystems we had to change from using udp to tcp for
> > nfs. I think we also bumped-up the rsize and wsize to something like 16384.
>
> But I also have problems with TCP connections (I assume scp uses TCP
> and not UDP, right?) when jumbo frames are enabled: scp just gets
> stuck.
You need to make sure that your switch ports are configured to support
jumbo frames. Most switches require at least some configuration to
support jumbo frames on a given port. Most Cisco switches simply have
an "enable/disable jumbo" flag, but others require setting the lergest
MTU size. Are you sure your switch is properly configured?
Unfortunately, the way you would normally test this is by running
tracepath between the hosts but since tracepath is broken in RHEL3 this
makes troubleshooting this issue difficult.
If you're really adventurous you could try building the kernel with the
linux-2.4.21-ipsec.patch backed out. This is the patch that appears to
cause the breakage.
Later,
Tom
--
Taroon-list mailing list
Taroon-list@(protected)
http://www.redhat.com/mailman/listinfo/taroon-list
Earn $52 per hosting referral at Lunarpages.
|
|
|