Fragmentation issue in Cisco 3750 stack

dmalla · ‎10-13-2008

NetPros,

We have a stack of 3750 (5 switches). There are two servers (on Fa5/0/37 and Fa5/0/38) that need to communicate with eachother. The requirement is to send the packets with size 2048 (fragmentation allowed).

When we ping from one server to another server with packet size of 2048, we get packets loss. On sniffer trace, we see "fragment reassembly time exceeded".

But when we take the 5th switch from stack and run the same test, the ping goes through.

Can somebody explain they we see fragment reassembly issue in stack and not in standalone switch?

Report Inappropriate Content · ‎10-17-2008

The possible causes for this type of issue may be due to the following reasons:

Normal Memory Utilization

Memory Leaks

Memory Fragmentation

In order to determine if fragmentation occurred, issue the show memory summary command in order to compare the Largest and Free fields.

Fragmentation occurred if the number in the Largest field is much smaller than the number in the Free field. This is because the Largest field indicates the largest contiguous free memory block and it should normally be close to the free memory.