The Oracle background process starts with the start of the instance, primarily maintaining the stability of the database, which is equivalent to a manager and internal service staff in an enterprise. They do not provide services directly to users. A: Database write--data write dbwr role: Synchronizes the modified data in the SGA to a disk file. Ensure that there is sufficient number of free blocks in the buffer cache. ps: If LGWR fails, DBWR will not obey the CKPT command strike, because Oracle writes the data buffer data to the disk before it writes the log buffers into the log file, and patiently waits for its first to complete, To complete this memory brush to disk action, which is called everything has a record. Trigger conditions: 1, Check Point ckpt 2, a service process did not find the free block within the set time 3, every three seconds automatically wake up once. Settings: Db_writer_process is used to define the number of dbwn processes. (The commit command simply writes the record modification to the log file, not the modified data to the data file) II: Log write:lgwr--logfile write (worker, very important busy process) function: Log The log contents in buffer are written to the online log file, releasing the log user buffer space. Trigger conditions: 1, the user issued a commit command. (called Fast commit in Oracle): Writes a record in the redo log buffer to a log file, writes a committed record 2, three seconds wake on time. 3, log buffers more than 1/3 log buffer, or the number of logs exceeds 1M. 4, DBWR process trigger:D bwn View writes a dirty chunk of data to disk first detects whether his related redo record is written to the online log file, and if not, notifies the LGWR process. Become an advance write mechanism in Oracle (write ahead): Redo record is written to disk before data record 5, online log file switching will also trigger LGWR. Three: checkpoint:ckpt-checkpoint event role: Maintain database consistency status. Point-in-time data files are consistent with the content in the SGA, which is not a separate process to work with the first two processes. DBWR writes dirty data while triggering the LGWR process. CKPT update the checkpoint record in the control file. Control the trigger time of the CKPT by setting a parameter adjustment. The parameter is fast START MTTR TARGET. Trigger Condition: Log switch triggers a checkpoint. Four: Process monitor:pmon--maintains user process Process monitor role: 1, discovers that the user process terminates abnormally and cleans up. Frees up resource usage. (Clean up the lock that is used by the terminating user) 2, the registered instance to the listener dynamically. TRIGGER Condition: The timer is awakened, and the other process will wake it up actively. V: System monitor:smon--Instance maintenance process Systems monitor role: 1, responsible for instance recovery, roll Forward back to the status of the instance shutdown, Redo using the log process after the last checkpoint. This includes both committed and uncommitted transactions. Open the database for rollback (roll back): Rollback of uncommitted transactions. (The data is not lost after Oracle commits commit, and we now have a general idea of how to implement this commitment, as well as a balanced selection between data security and database performance.) ) 2, responsible for cleaning up the temporary segment to free up space trigger conditions: Periodically wake up or be actively awakened by other transactions. VI: LCKN applies only to RAC databases and can have up to 10 processes (LCK0,LCK1,...,LCK9) for blocking between instances. Seven: ARchive:arch--Archive Operations role: Copy the full online log file to the archive directory when the log switch occurs. (LGWR write log write to overwrite overrides, trigger the arch process to transfer log files, copy out to form an archived log file, so as not to lose the log) trigger conditions: Log switching is LGWR wake. Settings: Log_archive_max_processes can set the number of arch when Oracle starts. Eight: Distributed database recovery --reco for the recovery of a distributed database an application that spans multiple databases requires that both commits succeed, the transaction succeeds, or all is rolled back. Above is the eight most core, most important, most useful process, and some of these processes are listed below. Nine: Manageability Monitor:mmon--awr main process role: 1, collect the data required by AWR, write statistics to disk. Saved in the Sysaux table space in 10g. 2, generate server--generated alarm hourly writes statistics from the shared pool to disk, or the shared pool occupies more than 15%. 10: Manageability Monitor LIGHT:MMNL: Lightweight mmon 11: Memory Manager:mman: Automatic content Management role: Check awr performance information every minute, And based on this information to determine the optimal distribution of SGA components. Settings: Statistics_level: Statistics level SGA_TARGET:SGA Total size 12: Job Queue Coordinator: cjq0--Database Timer task 13: Recover writer:rvwr Role: Provides logging for flashback databases. Writes the front image of the data block to the log. 14: Change tarcking writer:ctwr action: Track data block changes, record block address to Change_tracking file. An incremental backup of Rman will use this file to determine that the data blocks have changed and are backed up. above is a relatively common background process in Oracle, not all processes.
Oracle Background Process detailed
