The sp_cluster logical commands deactivate, failback, failover, and offline are asynchronous. They stop an online instance that may have existing transactions. These transactions must be handled before the instance can actually be taken offline or made inactive. As a consequence, these commands can be allowed to complete at a later time and context.
When you execute any of these commands, the target instance is placed in the time_wait state, and no longer accepts new connections.
Each asynchronous command provides three “wait” options for handling existing transactions. Values are:
wait – lets existing connections remain for a specified period of time, for example five minutes. Connections that support migration migrate as soon as they are quiescent. Connections that do not support migration disconnect when they become quiescent. HA-aware clients fail over; clients that are not HA-aware disconnect. Connections remaining after the specified time are terminated.
until – lets existing connections remain until a specified time, for example 12:30 p.m. Otherwise, until and wait handle connections in the same way.
nowait – terminates existing connections immediately. Connections that support migration must migrate immediately, or they are terminated.
If you do not specify a wait option when executing an sp_cluster
logical asynchronous command, Adaptive Server assumes
an infinite wait.
When the last connection using the instance disconnects, the instance state changes from time_wait to offline or inactive.
