A long-running request starts when either a transaction begins, or when a request is made to invoke a stateful service. When in this mode, the connection between the client and the server is maintained until either the transaction ends, or the client indicates that stateful processing is no longer required. However, a transaction start request may occur when a long-running request is already in effect (for example, a request for a stateful request has been previously made), or when the first service request within a transaction is for a stateful service. In this situation, the end of transaction will not terminate the long-running request as the server is still running in stateful mode.
Once a long-running request has been initiated, the service execution process is only able to process requests made by this specific client. To maintain responsiveness to other clients, the enterprise server starts a new service execution process each time an existing service execution process starts processing a stateful request.
When stateful processing is in effect, the application container does not return to its initial state following successful execution of a service request. Instead it maintains the client state of the container across multiple service requests until it receives notification from the client indicating that stateful processing is no longer required. (For example, it does not cancel user programs; does not return environment variables, tunables, or switches back to their initial state; does not unload user libraries etc.) When it receives notification indicating that stateful processing is no longer required, the application container returns itself to its initial state.
Stateful mode is implicitly exited if a service request causes a run-time system error to occur.
Enterprise Server manages its service execution processes to maintain the configured number of processes. This may mean that the enterprise server terminates the service execution process when a stateful client request completes.