Nathan wrote:
> Update...
>
> Updating the Cisco switches to the latest IOS, locking the protocol as LACP,
> and setting both the switch and Solaris system to active LACP mode has at
> least improved connectivity to the point where some systems are now reachable
> over the trunked VLANs on the aggregated ports.
>
> Here's the Twilight Zone observation: for two hosts (one Sun, one linux box)
> that cannot ping each other over the connection, running snoop on the Sun box
> enabling the connection between the two hosts! As soon as snoop is closed,
> the connection jams again. Both hosts have arp cache entries for each
> other's mac addresses before running snoop.
>
This to me sounds very much like the unicast address is not properly set
on the nics on the Sun box. Running snoop disables the mac address
filter on the hardware (promiscuous mode), which means that if the
unicast address is not correct you'll still get packets.
> Another bizarre "trick" that seems to un-jam the connection is to run a
> constant ping and then unhook and reconnect one of the network cables. The
> cables and ports were all checked and appear fine.
>
That one is weird. But perhaps the loss of link is causing the driver
to do some internal reset... perhaps including reprogramming the address
filter in hardware?
At this point, I'd like to know what class of Sun hardware you are
using, down to the specific NIC driver in question.
-- Garrett
> I note this posting on comp.sys.unix that seems to describe a similar
> scenario:
>
> http://groups.google.com/group/comp.unix.solaris/browse_thread/thread/b0b0c7bf3c7a79e9/f5734aa1b1a6f7f4?lnk=gst&q=aggr&rnum=1#f5734aa1b1a6f7f4
>
> Any ideas?
>
>
> This message posted from opensolaris.org
> _______________________________________________
> networking-discuss mailing list
> [email protected]
>
_______________________________________________
networking-discuss mailing list
[email protected]
