Greg,
Your email was very helpful. I isolated the freebsd box, and the win2K
box on a LAN without connection to the upstream Cisco device, and all
the connectivity problems went away. I'll be implementing the 'turn off
STP' on those Cisco ports shortly.
Many thanks,
Matthew
Greg Barniskis wrote:
Matthew Pope wrote:
I find that during the blocking behaviour, when I try and ping the
windows box, a tcpdump shows that each second ping attempt is
followed by a response (it appears) from an IPv6 address...
13:30:51.066625 802.1d config 8000.00:30:19:53:05:00.8011 root
8000.00:30:19:53:05:00 pathcost 0 age 0 max 20 hello 2 fdelay 15
13:30:53.069431 802.1d config 8000.00:30:19:53:05:00.8011 root
8000.00:30:19:53:05:00 pathcost 0 age 0 max 20 hello 2 fdelay 15
If you're referring to the above samples as "appears from IPV6", those
are Spanning Tree Protocol packets originating from the Cisco switch,
and are unrelated to your ping test. You will see them on the wire
frequently even in the absence of any normal IP traffic.
You probably want the following Cisco configuration directive added to
those switch ports that do not connect the 2900 to other switches:
spanning-tree portfast
The presence of the STP packets may or may not be related to your
performance issues. They shouldn't be, but some buggy NICs/drivers do
seem to get freaked out by STP.
When STP is enabled on a switch port, it definitely will delay your
initial link establishment by 30 seconds or so, when the attached
computer is first powered up. That alone can confuse things when the
NIC is trying to negotiate a link speed and the switch is still
thinking about STP. It's even possible that you're getting a link
speed/duplex mismatch out of it, and of course that will play holy
hell with your response time.
_______________________________________________
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"
