In RDBMS, transactions will be concreted to log files the first, then periodically, the dirty data in memory for those committed transactions are written to data files. Such an operation is usually called checkpoint.
In SQL Server, this happens approximately every minute.
To explicitly request a checkpoint, CHECKPOINT command can be issued.
Other operations, such as database backup, will also trigger checkpoint.
To record checkpoint event in error log, trace flag 3502 can be enabled.
Checkpoint is captured in log file. It marks the start point of rolling forward operation (replay committed transactions) during database starting up. Uncimmitted transactions after the last checkpoint will be rolled back.
No comments:
Post a Comment