[OpenAFS] Network becomes terribly slow when cache manager flushes updates over xDSL

Harald Barth haba@kth.se
Tue, 07 Jul 2009 20:07:45 +0200 (CEST)

> I'll try this. I noticed by searching about this field the existence
> of a "jumbo" flag : I suppose it is not in official releases yet,
> since /sbin/afsd -help doesn't mention it? (and using it gives an
> error...)

There is the other behaviour of rx that is packing more than one
(small) AFS datagram into an rx packet.

There were some bugs associated with jumbograms, too. But setting
RX_MAX_FRAGS to 1 forces rx to not make UDP packets bigger than the
local MTU. If your local MTU is bigger than the MTU on your xDSL that
will not help, but probably you have at least an MTU of 1500 all the
way if no VPNs or IP-tunnels are involved.