We have some 3560G switches and 2950 switches. There is one 3560G used as a distribution switch connecting to other 3560G switches or 2950 switches. We configured 1000/full between 3560G connections and configured 100/full between 3560G and 2950 switches. And there is no link negotiation issue at all. But the weird thing is I found in the trunk interfaces of 3560G distribution switch to the other 2950 switches keep getting out-discard errors. And correspondingly the trunk interface in 2950 connecting 3560G distribution switch keep getting Rcv-Err. I checked the traffic through these trunk interfaces. Traffic there is very light. Less than 2Mbps. So we are not overusing the bandwidth which is 100Mbps.
The other Giga trunk links between 3560G distribution switch and the other 3560G switches has no error at all. I see chaning the hold queue size is a bad idea. Also I tried set up auto/auto at both ends and trimmed some traffic in the 100Mbps trunk links. Nothing helps. So if you have any experience on this, please help throw some light on this issue. I really appreciate it.
Normally auto/auto should be fine. When you have auto at both ends of the link, and you do a show int status, do you see "a-full a-100" at both ends?
After you set auto/auto, did you shut and no shut the link, and clear the counters?
Thanks Kevin for reply. "Show int status" shows 100-full. I configured as 100/full at both ends. I no shut and shut serveral times also cleared counters.
I just noticed those trunk interfaces from 3560G to 2950 switches have out-discard errors at the same time. I checked last week record in NMS, they all happened at the same time. It seems some same traffic like multicast traffic flowing through those trunk interfaces causing the out-discard. But the giga trunk interfaces has no problem at all. I need capture this traffic when errors happen next time. But it's really hard to capture them. They are bursty in nature.
Did you ever figure this out?
I'm seeing the same problem (OutDiscards with multicast traffic on 100mbit ports) with differing kinds of switches (mostly 2960 but also 2940 and 3560).
I actually came upon this while we were setting up a multicast stream. The stream (2,5mbps) is really choppy/bad for users with 100mbit uplinks but if we move the same user to a gigabit link it works perfectly. If any of the uplinks between switches somewhere in the multicast stream is a 100mbit port we the end result is really bad. Setting a gig port to 100mbit also causes the problem with OutDiscards on the port. If i I stop the multicast stream then there is no problem.
All the ports are under very light load (typicall 2-3mbps) so this appears to be directly caused by the multicast stream itself...