Steve Holdoway wrote:
On Mon, 23 Jun 2008 15:38:06 +1200
Roy Britten <[EMAIL PROTECTED]> wrote:
While perusing the MySQL Documentation
http://dev.mysql.com/doc/refman/6.0/en/communication-errors.html I
came across reference to a "Linux duplex syndrome". The writer
complains "Many Linux Ethernet drivers have this bug". Would anyone
care to comment and/or help me understand what they're on about?
Thanks,
Roy.
I hope this is long sorted now, but there was a time where the Cisco IOS /
catalyst switches wouldn't autonegociate properly - and end up in a state where
performance was absolutely terrible. This wasn't limited to linux servers
though - sounds a bit FUDdish to me - as I had the same problems on slowlaris
as well. IIRC, we're talking 2000/2001 here.
We just used to force the switch ports to full duplex, and do the same with the ethernet ports on the servers.
Steve
That was because Sun, in all their wisdom, decided to prioritise
100BaseT4 ahead of 100BaseT in the negotiation order.
Cisco talks 100BaseT4 as well so they set themselves up happily on that
and, as a result, didn't chat to anyone else. (or to each other that
well as 100BaseT4 users 4 pairs to talk rather than the now-standard 2
pairs of 100BaseT).
When you forced the negotiation you forced 100BaseT-FullDuplex and
surprise, surprise, it all worked! :-)
Brat.
