Solved: UCS server for CUCM/UnityCnx/UCCX retransmission issue

MARK BAKER · ‎09-02-2011

I have installed CUCM, UnityCnx, and UCCX using VMware 4.1 on a C210M UCS server. I thought the web interface was a little slower to change between menus than the one we have on a physical server, but didn't think it was too bad. The real issue was found when I tried to upgrade the 3 UC guest systems. I see continuous retransmissions between the SFTP server and the CUCM, UnityCnx, and UCCX servers. It takes a very long time to upgrade the systems because of the retransmissions. Has anyone else seen this issue?

DAVE GENTON · ‎09-02-2011

Not specific to retransmissions, no. But they are just an indicator of another issue somewhere. Not sure what you have, never seen your environment, dont know you from Adam. But if you are running 4.1 with UC guest, then you need to be aware of the TCP LRO issue if you are not already. LRO is enabled by default and it is fixed now with a new patch, but in 4.1 prior to the fix will cause extensive "TCP" issues usually in the form of poor TCP performace as a blanket statement. This can of course cause retransmissions among many other things. Here is the link to disable LRO on our guests and restore TCP transmissions back to normal, shall this be affecting you in the first place. You didnt mention you fixed known LRO issue prior to stating a tcp issue so I'll post it on assumption being you specifically called out a protocol that's tcp based and no udp based.

Good Luck, Let me know if it takes care of it.

http://docwiki.cisco.com/wiki/Disable_LRO

View solution in original post

DAVE GENTON · ‎09-02-2011

Not specific to retransmissions, no. But they are just an indicator of another issue somewhere. Not sure what you have, never seen your environment, dont know you from Adam. But if you are running 4.1 with UC guest, then you need to be aware of the TCP LRO issue if you are not already. LRO is enabled by default and it is fixed now with a new patch, but in 4.1 prior to the fix will cause extensive "TCP" issues usually in the form of poor TCP performace as a blanket statement. This can of course cause retransmissions among many other things. Here is the link to disable LRO on our guests and restore TCP transmissions back to normal, shall this be affecting you in the first place. You didnt mention you fixed known LRO issue prior to stating a tcp issue so I'll post it on assumption being you specifically called out a protocol that's tcp based and no udp based.

Good Luck, Let me know if it takes care of it.

http://docwiki.cisco.com/wiki/Disable_LRO

MARK BAKER · ‎09-07-2011

Dave,

We were actually using adapter type Flex. I'm not sure how that got set for Flex. We used Cisco's OVA file to create the quest so I assume that is were it was set or maybe just the default for VMWare 4.1. We changed to vmxnet3 and still had issues. We then applied the fix that you mentioned and everything started working. Thanks for your help on this.

Mark

DAVE GENTON · ‎09-07-2011

Glad to help Mark !