[Spread-users] Tuning Windows XP for Spread

[Doug Palmer]

On Thu, 2007-07-12 at 12:39 -0400, John Lane Schultz wrote:

Thanks for the explanation of s_retrans/u_retrans.

> If you are consistently getting tok_hurry's, which do occur by default every two 
> seconds, 

Does this mean that I should expect to see tok_hurry go up normally or
that, if there is something wrong, the tok_hurrys occur every 2s?

> lots of s_retrans and u_retrans then you probably have faulty network 
> hardware somewhere in your setup and packets are getting dropped by the network.

I'm reasonably sure that that is what we don't have.

I've done extensive testing, using spsend/sprecv, of various network
configurations and, provided that the receiver is not loaded, then all
the packets get through just fine. The fault appears to be with
something about the receiver. I've tried two different NICs with similar
results. So I suspect an OS/software problem. (I'll try using a
different switch today, just to see what happens.)

A colleague has suggested that http://support.microsoft.com/kb/233401
may be a problem. So I'll also look into that.

I can reliably trigger the s_retrans runaway by sending a message over
1k in size. So I suspect that the fragmentation of the message, possibly
in combination with tok_hurry getting all excited, is causing the
runaway. Below 1k, I get the occasional s_retrans, but nothing to take
exception to.