nexus 7000 otv issues

[bob brennan]

hi,

i have 2 nexus 7000's using otv. one is located in our data center and one is located at our dr site. we are running across a couple of issue that we have not been able to resolve.

the first issue:

duplicate ip addresses on extended vlans. we have only noticed this issue with windows 2008 servers. 2000 and 2003 work fine. if i have two 2008 servers in the same site and try to configure duplicate ip addresess w2008 can determine this and won't let you do it. if they are at different sites i can configure duplicate ip addresses and responses to ping's is alternated between servers. we ran across this when we had a confiugration error on one of our servers. has anyone else seen this ?

i turned off arp suppression and i still had the same issue

2nd issue

transfering files over otv from one server to another from the dr site to the data center works fine. but when i go from the data center to the dr site it is 20 times slower. same servers over the same network path. i did a sh hardware rate-limiter and i noticed that the layer-3 mtu drops were rising at the data center. i checked the policy-maps and the interfaces and i am seeing no errors. i am not sure what causes this counter to increase. does anyone have an idea or have they ran across a similar issue ?

thanks fgor your help.

[Jeff Lande]

first issue, sorry no clue.

second issue, do you have the MTU size set to 9216 on the L3 join interface?  or better yet, on all interfaces that can possibly be in the path from the datacenter to the dr site?

[bob brennan]

hey jeff,

thanks for you reply. i have all interfaces set to 9216 including join-interface. i have checked this numerous times. any other ideas ?

[bob brennan]

i do have a 3750 which only supports and mtu of 1998

in the data path. besides that everything else is 9216.

[bob brennan]

i opened a case with cisco tac and here are the results

Hi Bob,

From the troubleshooting, it appears there was an issue with the datacenter Nexus.  We checked the entire OTV path and the mtu settings were above 9000 for the path but on the prod nexus we were seeing drops on the hardware rate limit for mtu.  The data transfer was sending packets of size 1500 but was hitting the Nexus hardware rate-limiter which caused the massive delay in the data transfer.  Upon checking the programming in the Nexus, it looks like somehow this information was misprogrammed on the switch and the Nexus was not utilizing the actual mtu of 9216, therefore causing the problem.

After reloading the vdc on the problem Nexus, it looks like it cleared the misprogramming and we no longer have any transfer delays.

the issue was a software bug and the only way to fix it was to reload the vdc.