MTU on F5 LTM

Hi all,

 

We have 2 data center A and B; and there is one pair LTM and GTM in each DC. GTM provide global LB.

 

 

Our objective is active-active solution. A and B provide service at the same time.

 

 

We have realized one potential issue:

 

 

for one user from Internet who is accessing to Data Center A service, GTM may point him to Data center B if he is online more than 30mins (default DNS cache time), which is not what we want.

 

 

 

To fix this issue,we are going to build GRE tunnel between 2 data centers LTMs to provide site affinity so that the customer traffic can be send back from Data Center B to Data center A by LTM using cookie persistence check.

 

 

However, the above solution introduces a MTU issue for us. Our Data center infrastructure has not introduced Jumbo frame support so MTU on all the boxes is 1500bytes.

 

 

Accordingtomyunderstanding, as GRE tunnel introduce the overhead to the original packet and DF bit is set by default, the "big" size packet w ill be dropped by GRE tunnel. Or ltm can do fragment before packet goes into gre tunnel. i dont know which idea is correct?

 

But it looks like that LTM doesn't support Jumbo frame on my box 10.2.2 so that i can't increase MTU for GRE tunnel interface on F5 LTM to accommodate the big size packet+GRE overhead.

 

 

Any suggestion for my design scenario? Your help is highly appreciated.