Oracle background process function description

Oracle background process function description

1. SMON: System Supervision Process
1. It mainly refers to the submitted data, but the data file has not been written to the power-off, and the smon will be restored upon restart.
A. Perform rollback to write data that has been written to the redo log file but has not been written to the data file (using the scn number to identify the submitted Records)
B. Open the database
C. Roll Back uncommitted transactions
2. You can also perform disk space maintenance.
A. Reclaim or merge the idle zones connected to data files
B. Release temporary segments
Ii. PMON: process supervision process
1. When a process crashes, PMON will be responsible for the following cleaning work:
A. roll back the current user transaction
B. Release all table-level and row-level locks added by the user.
C. release all other resources of the user.
Iii. CKPT: Check the process of the checkpoint and checkpoint
1. Improve system efficiency and Database Consistency, and introduce verification point events. CKPT is used when DBWR writes data in the high-speed buffer to the data file.
In theory, it is not required.
The redo log and scn can be used to ensure full recovery. It is introduced to improve efficiency, because all changes so far from verification points have been written to the number.
The data file is in.
The redo log before the verification point is no longer needed, so that instance recovery is accelerated, CKPT will write the verification point to the file headers of all relevant data files, the checkpoint number, redo log serial number, archive log name, and maximum scn number are also written into the control file. Because ckpt produces a large number of IO operations, therefore, verification points cannot be generated frequently. It is best to take over 20 minutes.
4. ARCn: log archiving process
1. When a group of redo log files is full, they will be written into another group. When the last group is full, the first group will be written, which is called Log switching.
2. If archive is not enabled, it cannot be restored after overwriting.
3. If archive is enabled, the arch process will be used to write the archive log file. Oracle will ensure that the archiving of a group of redo logs is complete.
This group of logs will not be reused before
V. DBWn: Database write process
Before 1.11g, only 0 ~ 9, 10 processes, 36 after 11g, 0 ~ 9 and ~ Z. Use db_writer_processes to set the number of instances.
If the number is not specified at startup
This parameter is determined by the number of CPUs.
2. When the following conditions occur, DBWR will write data in the high-speed buffer to the data file.
A. When the number of dirty buffers exceeds the set limit
B. The specified time interval has reached
C. When a process requires a high-speed buffer, but there is no idle high-speed buffer space
D. Check when a checkpoint occurs
E. A table is deleted or truncated (truncate, drop)
F. A tablespace is set to read-only.
G. When online backup is performed for tablespace
H. When the tablespace is set to offline
Vi. LGWR: redo the log write process
1. Write the record sequence of the redo log buffer to the redo log file. Note that it is written in sequence, so IO is relatively small.
2. How the commit statement works. The Quick commit technique is used here.
A. the server process writes the submitted records together with the generated scn number to the redo log buffer.
B. LGWR writes records in the redo log buffer to the redo log file to ensure recovery.
C. oracle notifies the user (process) that the submission has been completed
E. The server process will modify the status of relevant data in the database high-speed buffer, release the resource, and open the lock (this data may not be written at this time ).
To the data file, they are marked as dirty data, and then written to the data file by DBWR)
3. In the following cases, LGWR will write the redo log file
A. transaction commit
B. Redo logs in the log buffer exceeds 1/3 of the buffer capacity.
C. DBWR write data blocks in the database high-speed buffer to the data file
D. Every 3 seconds