1

We have multiple web servers behind a Netscaler load balancer. The (Win2008) servers each host an ASP.Net application (using iis 6.1) that are configured to use a central server running the .Net StateServer service for session sate management.

When we configure the load balancer to use "round robin" or "least busy" routing, the web application regularly crashes with an error indicating that something it expected to be in the session state was missing. However, it doesn't do it ALL the time - only on certain steps and then on about 75% of the attempts.

When we configure the load balancer to server persistence (so the user "sticks" on one server) the issue does not occur. (But this is not our desired running mode)

Things we have checked/done already:

  • Restart all the things
  • The machine keys are the same on all servers
  • There are no connectivity issues between the web servers and the state server
  • IIS is site names, paths and id's are the same on all web servers
  • No error is logged in the application logs

Does anyone have any suggestions on other things to check / possible causes?

NB. We have exactly the same set up in another environment - same type of load balancers, same web app, same config, same server setup... and it works fine. The only difference is a different version of VMWare Tools but can't see this being the cause?

Stelloy
  • 141
  • 6
  • 1
    My suggestion, if possible, perform a packet capture of some of the packets reaching the server in the working environment and do the same thing in the non-working environment. In this kind of situations only a really closer look reveals the issue. – Gabriel Talavera Dec 30 '13 at 12:05

1 Answers1

0

Months later, we've discovered the cause. Some (but not all) of the servers had been upgraded to a newer version of the .Net Framework.

Microsoft doesn't support different servers using the same StateServer with different versions of the framework, so when customers hopped between a server with 4.5.1 and one that had an earlier version they essentially lost their session and the application crashed.

If you encounter this issue I suggest you check which Framework is installed on each server - they must all be the same for them to work with the same StateServer.

Stelloy
  • 141
  • 6