Written by Robert Kettel
When you start to use Web Application Proxy Server (WAP) as a replacement for ISA, TMG or UAG and publish Active Sync through it, you might face a lot of Event 13007 warnings in the Microsoft-Windows-Web Application Proxy/Admin event log (I mean, really “a lot”). These are paired with various Event IDs 13006 (“Connection to the backend server failed. Error: (0x80072efe)”).
On the other hand, you don’t get any complaints from the user community. There does not seem to be an impact at all.
Where do these warnings come from? Do they impact our service and can we prevent them from showing up over and over again?
The main cause for the warning 13007 and 13006 is how devices with Exchange Active Sync (EAS) and Direct Push technology connect to Exchange.
Following this Microsoft TechNet article, “a mobile device that’s configured to synchronize with an Exchange 2013 server issues an HTTPS request to the server. This request is known as a PING. The request tells the server to notify the device if any items change in the next 15 minutes in any folder that’s configured to synchronize. Otherwise, the server should return an HTTP 200 OK message. The mobile device then stands by. The 15-minute time span is known as a heartbeat interval.”
With other words, there is a steady HTTPS session for 15 minutes between the EAS device and the Exchange backend which must be supported by all components taking part in the HTTPS session build, usually firewalls, load balancers and (in our case) the WAP servers which proxy the HTTPS session coming from the LBs to the Exchange backend. The long lasting session request is finally terminated by the Exchange server by posting an HTTP 200 message.
Now, looking at the default settings of our WAP servers we find a parameter which can have an influence on that behavior.
The default value for the InactiveTransactionTimeoutSec parameter is 300 (= 5 minutes). That means that if the accessing party does retrieve new responses from the backend service defined in the application settings for more than 5 minutes, the connection is identified as “timed out” and dropped by the WAP server.
From that perspective a connection to the Exchange Backend Service is timed out for the Web Application Proxy (causing a warning event 13007), when
a) the heartbeat interval of Direct Push is longer than the InactiveTransactionTimeoutSec
b) the session was not renewed by the device
c) there was nothing to synchronize in the first 300 seconds of the connection
However, if the device’s HTTPS session is dropped by the WAP server, it will automatically re-initiate a new session (the same what the device would do when getting an HTTP 200 OK message from the backend Exchange server). Therefore this is not a critical behavior at all.
The difference can be found how the device reacts on the dropped connection in comparison with how it handles the HTTP 200 OK post.
In the latter case, the device starts a new HTTPS session with the same heartbeat interval again.
In the first case, the device “assumes” that 15-minute HTTPS requests are blocked and re-initiates a new session with only 8 minutes (480 seconds) – which is still beyond the 300 second default setting of the WAP servers.
To avoid the Event 13007, the InactiveTransactionTimeOutSec parameter needs to be set to a value greater than the defined Active Sync max period. By default, the Active Sync device starts with a 15-minute interval, which would fit into a value of 910 for the InactiveTransactionTimeoutSec.
Since the parameter can be set for each published application individually, you luckily do not need to touch other published applications than Exchange Active Sync.
Get-WebApplicationProxyApplication ExchangeActiveSync | Set-WebApplicationProxyApplication –InactiveTransactionsTimeoutSec
A support engineer with Microsoft Premier stated that there is currently no way to suppress the Event 13007 from appearing in the log files completely.