CKPT process Work
2.CKPT Process Working mechanism
The condition that the checkpoint process is triggered is:
A> when a log group switch occurs;
b> when a user commits a transaction (commit);
c> Redo Log buffer capacity reaches 1/3 or 1M of total capacity.
d> manual alter System checkpoint.
When the e> system shuts down properly.
F> other (such as Alter TABLESPACE. Begin Backup/end Backup)
When a checkpoint process occurs (Figure 2), the checkpoint SCN corresponding to the checkpoint is recorded first, and the latest redo byte address (Redo byte address:rba) for the log file corresponding to the DB buffer modified at that moment is recorded. Then wake up the DBWN process, the DBWN process is awakened will check the checkpoint queue (in fact, in the data buffer cache buffer in the dirty data list), the Redo byte address (RBA) before the dirty data written to the disk file (Figure 4), before writing, Checks if the redo information before RBA has been written to the online log file, and if not, the DBWN process wakes up the LGWR process to write RBA information from redo to the online log file (Figure 5), then dwbn to write to the disk file and update the data block SCN. It is also said that before the DBWN process writes dirty data to the disk, it must ensure that the log information corresponding to the dirty data has been written to the disk, and if the log information for dirty data is not written to the disk, the DBWN process wakes the LGWR process to write the log information and continues to write dirty data to disk after completion. When the DBWN process writes all the dirty data before RBA to the disk file, the checkpoint process updates the SCN (Figure 3) of the data file header and updates the SCN and RBA information in the control file (Figure 1).
[Online lesson excerpt] 2.CKPT Process Working mechanism