Exchange 2010 Addressbook Errors with NetScaler 11.0

I'm working with a client that is upgrading the load balancing infrastructure that services their Exchange 2010 organization. They are implementing NetScaler SDX appliances that run NetScaler VPX virtual load balancers. Being as this is new infrastructure, version 11.0 (first release June 30, 2015, latest release Oct 8, 2015) is being implemented.

To test the new load balancing, we modified the hosts file on few computers to direct the load balanced names (outlook.domain.com/webmail.domain.com) to the IP address of the new load balancer. Initially, this seemed to be working fine, but we didn't do any specific testing.

After a bit, we noticed that we often got the following error in Outlook when accessing the address book:

The connection to Microsoft Exchange is unavailable. Outlook must be online or connected to complete this action.

This error was reproducible after not accessing the address book for a few minutes. In addition, when looking at the connection status, we would see failures talking to the directory. Connectivity to mail generally seemed fine, but also had occasional failures.

Connection Status with directory access failures

When you research this problem, you will find lots of references to incorrect network timeouts closing connections before the application is ready. This is exactly what we seemed to be having. However, we adjusted every possible setting on the NetScaler to allow connections to last up to two hours. No matter what we adjusted, the same problem remained.

Other network components were the same as before. So, it seemed the problem had to be the new NetScaler. Clients still using the existing load balancer did not have any issues.

After adjusting everything we could think of, we called Citrix support. In conjunction with support, the network team did some packet traces and confirmed that resets were being performed on the RPC connections for the address book.

The fix ended up being using version 10.5 (latest release Sept 2015) instead of version 11.0 for the VPX instance. As soon as version 10.5 was implemented the mysterious reset issue was resolved and the response time for the directory lookups dropped significantly.

Connection Status after implementing version 10.5

Citrix is continuing to evaluate the issue and I'll update when I have more information.

Update (Nov 10/15): Apparently NetScaler 11.0 has issues with Outlook Anywhere also: http://discussions.citrix.com/topic/368750-netscaler-11-6210-outlook-anywhere-broken/page-2

Update (Nov 13/15): Running version 10.5 we've still had no issues. It has been submitted as a bug with Citrix and we're waiting for a update. Hopefully the next revision will fix it, but they are definitely aware it's an issue.

Update (Nov 27/15): Citrix did some testing with the network folks again today with Version 11.0 (I'm assuming a new build) and still not working properly. We are back to Version 10.5.

Resources:

• NetScaler Release Software Dates: http://support.citrix.com/article/CTX121840
• http://blogs.technet.com/b/david231/archive/2015/03/30/for-exchange-2010-and-2013-do-this-before-calling-microsoft.aspx
• https://discussions.citrix.com/topic/348704-exchange-2010-rpc-timeout-issue/
• https://social.technet.microsoft.com/Forums/exchange/en-US/e7c5ab7c-7ccc-4f4a-ac40-31831aea2d81/outlook-time-out-setting?forum=exchange2010
• https://masterxen.wordpress.com/2015/08/11/exchange-rpc-timeouts-when-using-citrix-netscaler/