Intelligent Session Retry

The Intelligent Session Retry feature provides the ability to manage session retries when a connection to a specific node is down so that all Processes are not retried at the same time. This feature makes the existing connection retry facility more efficient by reducing the extra processing overhead created when each IBM® Connect:Direct® Process retries the connection independently.

When multiple Processes are submitted for a specific node that is down, the Processes all initially attempt to connect. When this fails, the first Process submitted to this node continues to retry until MAXRETRIES is exceeded. The other Processes to this node are moved to the Hold (HO) queue with a RH (held for retry) status and are not retried. When MAXRETRIES is exceeded for the first Process, it is moved to the Hold (HO) queue with a WC (waiting connection) status for session retry.

After the first Process is moved to HO WC status, it can be restarted by one of the following ways:

  • You can manually release the Process causing it to be retried until a connection is made or MAXRETRIES is exceeded, when it will return to HO WC status
  • Another Process is submitted to this same node, causing the Process in HO WC status to be retried until either a connection is made or MAXRETRIES is exceeded, returning it to HO WC status
  • The SNODE initiates communication with the PNODE, causing the Process in the HO WC status to be released

When the first Process in HO WC status connects and is released, this causes the other Processes being held in HO RH status to be released serially. Therefore, only one Process going to an inactive node will retry until the connection is made resulting in less overhead, since multiple Processes going to the same node are not all attempting to retry at the same time. This also reduces the number of statistic records which would be produced for multiple retries.

Note: The node name used by the Intelligent Retry feature is the specific name used for the SNODE parameter. This means that if the SNODE is known by more than one node name or TCPNAME value, only those in the queue waiting with that value will be recognized for release.