Flow control prevents session manager from exhausting critical system resources, which may occur under extreme workload.
Flow control does the following:
Monitors the status of system resources for session manager:
Raises events when a certain threshold has been reached:
PROACTIVE - Gives early warning to system components that can make memory available when required
SEVERE - Starts to scavenge as much memory as possible, current clients work fine
CRITICAL -Starts to slow data inflow to the session manager and raise the priority of getting data out of the session manager. Rejects new connections, suspends new sessions from currently attached clients, and pends new tasks in those suspended sessions—current sessions and tasks work fine.
HALT - Session manager enters into lockdown mode, no further processing is allowed until an administrative action is performed, or the system enters a safer state.
If the SSM on Linux remains at a critical memory level when there is enough available memory and not much unfinished workload, the SSM may not be detecting correct memory usage. This can cause the boundary event not to be triggered properly. If this situation occurs, try setting the following environment variables for the SSM in your application profile:
To speed up paging and session manager recovery, the specified directory can be on the local drive since the paged data of non-recoverable sessions does not need to be persisted at a shared location.
The following elements and attributes in the SOAM section of the application profile are related to task message data and common data paging for non-recoverable sessions:
For detailed descriptions of these attributes, see the Platform Symphony Reference.