Checkpoints control recovery in both nonclustered Adaptive Server and Cluster Edition systems, but the process for which they supply the transaction information differs. In the Cluster Edition, the oldest active transaction found is the oldest active transaction across the entire cluster, and dirty buffers are flushed on all clustered instances.
In both types of processing, Adaptive Server finds the oldest active transaction information at the time the checkpoint record is logged, uses it as a cutoff point, and writes the checkpoint record to the log. The dirty buffers are then flushed, up to the current flush sequence number. This checkpoint record is registered as a recovery checkpoint only after all the dirty buffers are flushed.
