we are observing 0.01% and 0.12% InputErrors on an Etherchannel between a 6500 and a 4500 consisting of 4 (resp. 3) GigaBit Links. We discovered that the Packets were simply droped due to "Rx-No-Pkt-Buff" - We have a PRTG Graph showing a rather low utilization of 3% or less (5 min average) over the whole day or week. Customer experiences disconnect session in his SAP and Mainframe Aplications ...
Anybody any Idea how these errors could come into existance? The counters were cleared a few days ago to confirm this observation ...
The EtherChannel Load-Balancing is src-dst-ip (the default) - an the errors are distributed 15 - 30 % among the 4 ports.
Nothing Suspicious in the Logs - a few OutDiscards on some Ports/Channels to a range of servers.