Implementation technology of recovery

Key issues involved in recovery mechanism

1. How to build redundant data
Data dump （backup）
Register log files （logging）
2. How to use these redundant data to implement database recovery

Data dump

What is data dump

Dump means that the database administrator periodically copies the entire database to tape 、 The process of saving on disk or other storage medium
The backup data text is called the backup copy (backup) Or backup copy
After the database is destroyed, the backup copy can be reloaded
Reinstalling a backup copy can only restore the database to the state it was in when it was dumped
To recover to the state when the fault occurred , All update transactions since the dump must be rerun

Above picture ：
The system is in Ta Stop running transactions at any time , Perform database dump
stay Tb Time dump is complete , obtain Tb A consistent copy of the database at all times
The system runs to Tf Always break down
To restore the database , First, the database administrator reinstalls the backup copy of the database , Restore the database to Tb The state of the moment
Rerun from Tb～Tf All update transactions at any time , Restore the database to a consistent state before the failure

Dump method

Static dump
Dump operation when there is no running transaction in the system
The database is in a consistent state at the beginning of the dump
No access to the database is allowed during dump 、 To amend
The result must be a consistent copy of the data
advantage ： Implement a simple
shortcoming ： Reduced database availability
The dump must wait for the end of the running user transaction
The new transaction must wait for the dump to end
Dynamic dump
Dump operations are performed concurrently with user transactions
Allow access to or modification of the database during dump
advantage
Don't wait for a running user transaction to end
Does not affect the operation of new transactions
Disadvantages of dynamic dump
There is no guarantee that the data in the copy is correct and valid
Example: at a certain time during the dump period Tc, The system sends the data A=100 Dump to tape , And at the next moment Td, A transaction will A Change it to 200.
On the backup copy A Out of date
Use the replica obtained by dynamic dump for fault recovery
It is necessary to register the modification activities of each transaction to the database during dynamic dump , Create a log file
The backup copy plus the log file can restore the database to the correct state at a certain time
Mass dump : Dump all databases at a time
Incremental dump : Dump only the data updated since the last dump
Comparison between mass dump and incremental dump
From a recovery perspective , It is often more convenient to use the backup copy obtained from mass dump for recovery
If the database is large , Transaction processing is very frequent , Then incremental dump is more practical and effective
Dump method classification

Register log files

Format and content of log file

What is a log file
Log files (log file) It is used to record the update operation of transaction to database
Format of log file
Log files in units of records
Log files in blocks
Log file content in record units
Start mark of each transaction (BEGIN TRANSACTION)
The end tag of each transaction (COMMIT or ROLLBACK)
All update operations for each transaction
All of the above are recorded as a log in the log file (log record)
Log files in units of records , The content of each log record
Transaction ID （ Indicate which business ）
Operation type （ Insert 、 Delete or modify ）
Action object （ Record ID、Block NO.）
Old value of data before update （ For insertion operations , This item is null ）
The new value of the updated data （ For delete operations , This item is null ）
Log files in blocks , The content of each log record
Transaction ID
Updated data block

The role of log files

purpose
Perform transaction recovery
Carry out system fault recovery
Assist backup copy in media recovery
Specific role
Log files must be used for transaction recovery and system recovery .
Log files must be created in dynamic dump mode , Only by combining backup copy with log files can the database be recovered effectively .
In static dump mode , You can also create log files .
When the database is destroyed, the backup copy can be reloaded to restore the database to the correct state at the end of the dump
Using log files , Redo a completed transaction
Undo the unfinished transaction at the time of failure
The database can be restored to the correct state at a certain time before the failure without running those completed transaction programs again

Register log files

To ensure that the database is recoverable , Two principles must be followed when registering log files
The registration order is strictly in accordance with the time sequence of concurrent transactions
You have to write a log file first , Write database after
Write log file operation ： Write the log record indicating this modification to the log file
Write database operation ： Write the changes to the data into the database
Why write a log file first
Writing a database and a log file are two different operations
There may be a fault between these two operations
If the database modification is written first , This change is not registered in the log file , Then we can't restore this modification later
If you write a log first , But it didn't change the database , It is unnecessary to perform the log file recovery only once more UNDO operation , It doesn't affect the correctness of the database
Welcome to join me for wechat exchange and discussion （ Please note csdn Add ）