In Active/Active configurations, the SAP Primary Application Server Instance (PAS), ABAP SAP Central Services Instance (ASCS) or SAP Central Services Instance (SCS) and Database (DB) hierarchies are separate and are in service on different servers during normal operation. There are times, however, when both hierarchies will be in service on the same server such as when one of the servers is being taken down for maintenance. If both hierarchies are in service on one of the servers and both servers go down, then when the servers come back up, it is important that the database hierarchy come in service before the SAP hierarchy in-service operation times out. Since LifeKeeper brings hierarchies in service during startup serially, if it chooses to bring SAP up first, the database in-service operation will wait on the SAP in-service operation to complete and the SAP in-service operation will wait on the database to become available, which will never happen because the DB restore operation can only begin after the PAS, ASCS or SCS restore completes. This deadlock condition will exist until the PAS, ASCS or SCS restore operation times out. (Note: SAP will time out and fail after 10 minutes.)
To prevent this deadlock scenario, it is important for this configuration to set the switchback flag for both hierarchies to Automatic Switchback. This will force LifeKeeper to restore each hierarchy on its highest priority server during LifeKeeper startup, which in this case is two different servers. Since LifeKeeper restore operations on different servers can occur in parallel, the deadlock condition is prevented.
Post your comment on this topic.