Re: HELP: Thin Clients freeze after 30 minutes.

[Ralf <rgx@gmx.de>]

Hi there,

problem was hopefully solved - after some 7 hours of collecting 
experiences....

Am Dienstag, 20. April 2004 21:44 schrieb Ralf:
> 1. Swap NICs:
> - ifdown
> - swap eth0 and eth1 in /etc/network/interfaces
> - exchange CAT cables
> - ifup

this is not enough: etc/... -  mh, where exactly was it? - anyway: the 
INTERFACE variable for the dhcpd3 server and the dhcpclient-2.2 (or 
similair) had to be exchanged, too, in the corresponding conf files.
>
> 2. Monitor traffic using nload

Traffic went never up linke before after netcard swapping.

> 3. Try to set a fixed 100Mbit full duplex mode on the NIC that serves
> the thin clients (mii-diag or eththool)

mii-tool -w ("watch") said that 100Mbit FD was established through 
autonegotiation - so this was no aparent problem. Therefore we changed 
the switch (10 Mbit with 100 Mbit uplink) and now everything went quite 
fast! We now were quite sure that it was the switch (however it could 
still be addressed from a windows machine - strange). But:
When I accidently had un- and repluged the ltsp server's plug from the 
back bone (internet / tjener), all thin clients froze at once. nload 
showed no traffic at all. Until, after some minutes of investigation, 
we entered
# mii-tool -A 100baseTx-FD
all thin clients popped up again simultaneously! So it WAS a 
flow-control problem despite mii-tool -w not reporting anything!

As a consequence, we inserted a pre-up line into interfaces like
pre-up mii-tool -A 100baseTx-FD,10baseT-FD

If this will do it, we can only hope...

Even more comments desired :)

regards
Ralf