Test #10 - Cluster: Single Application Server Abends
This demonstrates overall system behavior when the application server abends. Manually terminating the application server process while the application server cluster it is actively processing events from the active mediation server simulates the server abend. After approximately two minutes the application server restarts.
• The standby server discards any events received until the application server process ends.
• The active server successfully delivers any events received until the application server process ends.
• After a possible delay from the remote call to publish events timeout--during which events are queued by both the active and standby servers--processing resumes.
• The total number of events delivered to the application server cluster is 30,000, and the standby server drops all 30,000, less the number of events queued. Because the active server does not receive any events after the 30,000, the standby server should still have a batch queued which has not yet been dropped.
Actual Results: SUCCESS
• Active Server - The active server delivered 5400 events before the appserver process being terminated. Because of delivery delays for events, the active server queued events, and continued to queue for approximately three minutes, eventually queuing approximately 17500 events. The server then resumed delivering events, and eventually delivered all 30000.
• Standby Server - The standby server discarded 5200 events before the application server process ended. Due to the delay of delivery of events by the active server, the standby server began queuing events, and continued to queue for approximately three minutes, eventually queuing approximately 18000 events. The server then resumed discarding events, and eventually discarded 29400, with 600 remaining in batch.