3750 packet loss

Unanswered Question
Apr 4th, 2008

Hi, all.

We have a network of 4 3750. Connected sequentially by 100M Ethernet copper. Switches are mainly used to transport multicast data.

The problem is that large amount of data gets lost.

qos is disabled, duplex is set to static 100/full.

Investigation took us to asic dropping the packets. For example:

#show pl port-as stats drop gi1/0/23

Interface Gi1/0/23 TxQueue Drop Statistics

Queue 0

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

Queue 1

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

Queue 2

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 0

Queue 3

Weight 0 Frames 0

Weight 1 Frames 0

Weight 2 Frames 7598324

Why are the packets queued this weird? And how can we solve the problem ?

Please advice, we're out of ideas.

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Joseph W. Doherty Fri, 04/04/2008 - 16:41

Likely that's where the multicast traffic, and perhaps all traffic, is queuing and also overflowing the queue.

Since you mention the traffic is mainly multicast, there might not be a real advantage to try adjusting queues.

Perhaps the easiest thing to try is to increase the bandwidth between switches and or change their topology.

You mention "Connected sequentially by 100M Ethernet copper", if that's literally true, you might define one switch as the "root" and connect each of the three others to it.

You might also consider, if not used already, using your gig SFP ports with copper. Depending on model, you might also do this too in a star configuration.

You might consider stacking the units, if they're physically close enough.

You might consider defining an EtherChannel with two ports between switches (if you do this, check whether the default hash algorithm is acceptable, if not, change it.)

Actions

This Discussion