Whilst load balancing Exchange 2013 with a Capax Zantas EAS deployment in place I was reminded of an issue I experienced a few years ago with a Kemp load balancer. The EAS app for OWA was failing with various errors such as:

Communication with service failed. The service might be down or misconfigured. Error code: 500

As well as the search functionality never returning results.

And in the EASOWABackend log file we were seeing the following:

w3wp.exe Error: 0 : 22 : [09/23/2016 10:13:22.880 AM] User username@domain.com: EWS Call failed Microsoft.Exchange.WebServices.Data.ServiceRequestException: The request failed. The remote server returned an error: (501) Not Implemented. —> System.Net.WebException: The remote server returned an error: (501) Not Implemented.
   at System.Net.HttpWebRequest.GetResponse()
   at Microsoft.Exchange.WebServices.Data.EwsHttpWebRequest.Microsoft.Exchange.WebServices.Data.IEwsHttpWebRequest.GetResponse()
   at Microsoft.Exchange.WebServices.Data.ServiceRequestBase.GetEwsHttpWebResponse(IEwsHttpWebRequest request)
   — End of inner exception stack trace —
   at Microsoft.Exchange.WebServices.Data.ServiceRequestBase.GetEwsHttpWebResponse(IEwsHttpWebRequest request)
   at Microsoft.Exchange.WebServices.Data.ServiceRequestBase.ValidateAndEmitRequest(IEwsHttpWebRequest& request)
   at Microsoft.Exchange.WebServices.Data.MultiResponseServiceRequest`1.Execute()
   at Microsoft.Exchange.WebServices.Data.ExchangeService.InternalConvertIds(IEnumerable`1 ids, IdFormat destinationFormat, ServiceErrorHandling errorHandling)
   at Microsoft.Exchange.WebServices.Data.ExchangeService.ConvertId(AlternateIdBase id, IdFormat destinationFormat)
   at EASOWA2013.EASOWABackend.OpenMessage(String owaID, String userEmail, String authToken, String pageURL)

This error contained a 501 Not Implemented message, and this message was getting exposed in OWA. But this error was not present in any of the Exchange Server logs, so it was not being caused by Exchange. This led us to looking at the load balancer, and testing EAS when bypassing the load balancer. To bypass the load balancer requires host file changes and changing the EAS configuration, so it is not a trivial change. When the load balancer was out of the traffic path everything worked fine.

Under the Kemp system logging (Logging Options > System Log Files > System Message File) we were seeing:

Sep 22 16:39:09 cwh-kemp01 kernel: L7: badrequest-client_read [192.168.x.y:44987->192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501): 192.168.lb.vip:443] (-501):

Where 192.168.x.y was the client IP and 192.168.lb.vip was the load balanced IP for Exchange.

With reference to the above linked article, the fix was to change the Kemp’s handling of the 100-Continue feature and by changing the setting to RFC-7231 Compliant EAS in OWA started working: