Oracle launch three-phase

Startup Nomount

# #启动实例, start the first step, Oracle first look for the parameter file, and then according to the settings in the parameter file (memory allocation settings), create an instance, allocate memory, start a background process, nomount process is the process of starting a DB instance,

The first instance of the Spfile<oracle_sid> parameter file is determined and looked up according to Oracle_sid, and the process of fork on the operating system is created on the ORACLE instance when it is started, based on this oracle_sid.

Db_name is used to define the database name, must be a text string of no more than 8 characters, db_name represents the database name that the instance is about to mount, and during the creation of the database, Db_name is recorded in the data file, the control file and the log file. Also recorded in the SPFile parameter file, the parameter file must have at least the db_name parameter in the instance startup process, and the database cannot start if the db_name in the parameter file and the database name in the control file are different in the database startup Mount stage

ALTER DATABASE Mount

# #启动数据库到mount状态 Startup Mount

After booting to the Nomount state, Oracle can obtain the location information of the control file from the parameter file, locate the control file, and determine the existence of the data file based on the data file location recorded in the control file.

ALTER DATABASE Open

# #启动数据库到open状态

The control file records the data files in the database, the location information of log files, checkpoint information and other important information, in the database open phase, Oracle will find these files according to the information recorded in the control file, and then perform checkpoints and integrity checks, if there is no problem, you can start the database.

During the open process, the checks performed by Oracle include the following two items:

Check the checkpoint count (checkpoint CNT) in the header of the data file for the first time is consistent with the checkpoint count (checkpoint CNT) in the control file, which is checked to confirm that the data file is from the same version rather than recovering from the backup.

The second check of the data file header begins the SCN and the end SCN of the file that is recorded in the control file is consistent, if the ending SCN recorded in the control file equals the beginning SCN of the data file header, then the file does not need to be recovered.

This article from "Husband Minister" blog, declined reprint!

Oracle launch three-phase