I am beginning to use our load balancers in our new Exchange 2010 environment. I'm passing through RPC and MAPI and everything is working great.
However, it seems that the timeout value for the virtual server might be too low. I increased it from the default of 300 to 600 but that hasn't really helped. I'm considering disabling the timeout all together. Are there any risks associated with that? (excess burden on the load balancers?)
Solved! Go to Solution.
Certainly you have the ability of setting timeouts to long values or even disabling them, but I would suggest first understanding your workloads.
You might find there is no issue with current resources, or based on your usage, maybe increasing the ammount of memory allocated to your STM might be recommended. (incidentally, this is the *best* thing about softwware / virtual ADC's in my honest oppionion!!)
Can you advise how many users you are providing CAS load balancing for? It would also help to understand your system configuration - How many CPU's have you allocated and how much memory?
The purpose of the 'timeout' settings are to close a connection that has been idle for too long, on the assumption that the other end of the connection has stalled (or lost interest). The majority of TCP-based protocols are robust by design and will automatically reconnect when they need to submit another transaction (HTTP keepalive connections for example).
It's a good idea to close idle connections because they consume a (small) amount of resources on the load balancer - a file descriptor and a small amount of memory (often in the region of 10-20 Kb, but it varies by workload).
You mention that changing your timeout setting has no apparent effect. This may be because either the client or the stingray device can elect to close the connection; if the Stingray timeout is increased from 300 to 600 seconds, but the client has a timeout of 60 seconds, then connections will only stay idle for 60 seconds before the client closes them.