z/OS MVS Programming: Sysplex Services Guide
Previous topic | Next topic | Contents | Contact z/OS | Library | PDF


Handling Failed Connections During Rebuilding

z/OS MVS Programming: Sysplex Services Guide
SA23-1400-00

If a connection fails or disconnects abnormally (REASON=FAILURE) during rebuilding, the system frees any resources for the user and reports the failed event to the event exit of all connected users. Existing connections must develop protocols to determine if they should continue to rebuild the structure.

If the failed user specified CONDISP=KEEP at connect time, the connection becomes failed-persistent. Existing peer connections might also need to develop special processing to handle this situation.

In some instances, the system allows a peer connector to respond on behalf of a failed connector before all responses have been received. See Providing a Response for a Failed Connector. After all existing connections have responded to the failed event through the event exit, the system also handles any outstanding event response that the failed connection needed to provide.

When connections rely on each other to coordinate rebuilding, they must coordinate how to respond when one of them fails. For example, connection A and B are each responsible for completing the rebuilding of a structure. Connection A rebuilds its share of the data into the new structure, but connection B fails before it can rebuild its data into the structure. The following occurs:
  • Connection A responds to the failed event of connection B by issuing IXLEERSP with EVENT=DISCFAILCONN. Connection B has CONDISP=DELETE and is deleted.
  • The system reports a Rebuild Cleanup event to the event exit of connection A.

At this point, connection A cannot stop the rebuilding process, and the new structure does not contain data updates from connection B.

To avoid this scenario, connection A can
  • Stop the rebuild process prior to responding to the disconnect/failed event
  • Issue the IXLEERSP response to delete connection B
  • Perform recovery for connection B
  • Initiate® another rebuilding operation

If connection A is able to perform processing for connection B, connection A could also complete rebuilding the structure and then issue IXLEERSP to respond to the failed event. Thus, the structure can be rebuilt with the necessary data.

If all connections to the structure fail prior to the Rebuild Cleanup phase, the rebuild is stopped and the new structure is deallocated. If all connections fail during the Rebuild Cleanup phase, the rebuild is completed and the old structure is deallocated. ENF event code 35 is issued in either case when the structure is deallocated.

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014