918868 wrote:Imagine a session launches an UPDATE to a table with MANY, Many, many rows and every row gets changed.
Can anyone explain the 4th scenario i.e "Thee database must advance the checkpoint, which is the position in the redo thread from which instance recovery must begin. " under which DBWR writes into data files ?
918868 wrote:I am not sure that your understanding is correct about how oracle db works. The flow is,
Final clarification, sorry I should have asked in last post itself, before writing those dirty blocks to data files, checkpoint occurs and New SCN s generated and that SCN is updated in datafiles header and control files, hence the database should advance the checkpoint to incorporate the new checkpoimt information ie checkpoint time stamp and new SCN number generated ?
Edited by: 918868 on Jan 31, 2013 9:04 PM
918868 wrote:what will you do differently after you understand sentence above?
As written already about one of the scenario's when DBWR writes as per documentation
"The database must advance the checkpoint, which is the position in the redo thread from which instance recovery must begin."
What does this mean ?
918868 wrote:There are many different types of "checkpoints" (see http://jonathanlewis.wordpress.com/2007/04/12/log-file-switch/ ) The two best known are the log file switch checkpoint, and the incremental checkpoint.
Why does database advance the checkpoint? Aman told in last post that Oracle did not advance the checkpoint due to SCN generation/updates so it is due to other reasons? What is the relation to this with buffer writes with the advance of checkpoint?