Output Drops on 1gb and 10gb Interfaces

jrabener · ‎10-30-2009

I'm seeing many interface both 1gb and 10gb and EtherChannels with large numbers of output drops being logged.

The utilization on these interfaces (particularly 10gb) is relatively low.

What might be causing the drops?

How might I troubleshoot/resolve this situation.

Thanks,

Joe

Joseph W. Doherty · ‎10-30-2009

"What might be causing the drops?"

Insufficient ouput queue(s) or buffers to handle traffic burst.

"How might I troubleshoot/resolve this situation."

If excessive drops is caused by insufficent resources, increase resource allocation, if possible, may mitigate issue.

BTW, traffic bursting isn't easily "seen" especially with longer term utilization measurements.

Also BTW, excessive drops on TCP traffic will drive down average bandwidth utilization if TCP flows need to wait on timed out ACKs or are forced back to slow start where flow's send window never really enlarges.

jrabener · ‎10-30-2009

Thanks for your reply.

Giuseppe Larosa · ‎10-31-2009

Hello Joe,

if in the recent past a bridging loop caused a broadcast storm in the campus you will see high output drops counts also on 10giga links.

if these counters are not incremeting over time you are fine.

I've noticed this in our campus networks on C4500 and on C6500 devices.

I had thought this could be the result of broadcast storm-control that we use at 1% but the feature is declared to be inbound not outbound.

However, this has happened few times and everytime after a loop issue I look at counters I see these high output drops counters.

The key point is to see if these output drops don't increment over time.

if they increment over time other considerations like the ones done by Joseph apply to your case.

Hope to help

Giuseppe