06. on several ways of redis persistence

Personal home page ： Programmer a Hong
Ordinary life , The heart will shine , Years are dreary , Run , There will be wind
Recommended articles ：
【 Graduation season · The technique of attack er】 A year after graduation
Java A one-stop version of the learning route
A detailed explanation SpringBoot Global profile

1. RDB

Redis DataBase

• Within a specified time interval , Writes a snapshot of a dataset in memory to disk ;
• The default is saved in /usr/local/bin in , file name dump.rdb;

1.1 Automatic backup

• redis Is a memory database , Every time we run out redis, close linux when , Logically speaking , Memory free ,redis The data in will also disappear
• Why do we start again redis When , Yesterday's data is still , It didn't disappear ？
• Precisely because , Every time you shut down ,redis The data will be automatically backed up to a file ：/usr/local/bin/dump.rdb
• Next, let's have a comprehensive understanding Automatic backup mechanism

1. The default automatic backup strategy is not conducive to our testing , So modify redis.conf Automatic backup policy in files

vim redis.conf
/SNAP  #  Search for 
save 900 1   # 900 Seconds , At least change 1 Time , Will automatically backup 
save 120 10 # 120 Seconds , At least change 10 Time , Will automatically backup 
save 60 10000 # 60 Seconds , At least change 10000 Time , Will automatically backup 

Of course, if you just use Redis Cache function of , No need to persist , Then you can comment out all the save Line to disable the save function . You can use an empty string to disable ：save “”

2. Use shutdown Analog shutdown , Before and after shutdown , contrast dump.rdb File update time

Be careful ： When we use shutdown command ,redis The database will be backed up automatically , therefore ,dump.rdb The file creation time has been updated

3. Boot up redis, We will be having 120 Save in seconds 10 Data , Check it again dump.rdb File update time （ Open two terminals
   window , Convenient view ）
4. 120 Save in seconds 10 The action of data triggers the backup instruction , at present ,dump.rdb In the file 10 Data , take
   dump.rdb Copy a dump10.rdb, At this point, both files are saved 10 Data
5. Now that the data has been backed up , Then we will wantonly delete all the data flushall, Again shutdown To turn it off
6. Start again redis, Found that the data really disappeared , Not as we thought take dump.rdb Restore the contents of the file to redis in . Why? ？

because , When we save 10 When more than one piece of data , The data is backed up ;
Then delete the database , Back up the data in the file , Also no problem ;

however , The problem is shutdown On , Once this command is executed , It will be backed up immediately , Generate a backup file for the deleted empty database , Put the front into 10 The backup file of data is overwritten . therefore , The result of the above figure appears . Automatically
Recovery failed .

How to solve this problem ？ Back up the backup file again

7. take dump.rdb File deletion , take dump10.rdb Rename it to dump.rdb
8. start-up redis service , Sign in redis, data 10 strip , All recovered ！

1.2 Manual backup

• Automatic backup before , A lot of data has to be changed , For example, above , We changed more than ten pieces of data , Will automatically backup ;
• Now? , I only save one piece of data , Just want to back up immediately , What to do ？
• Each operation is completed , Carry out orders save It will be backed up immediately

1.3 And RDB Related configuration

• stop-writes-on-bgsave-error： Water inlet and outlet , Whether the water outlet fails or not
  • yes： When the background backup is wrong , The foreground stops writing
  • no： Dead or alive , It's going in
• rdbcompression： For snapshots stored on disk , Whether to start LZF Compression algorithm , It usually starts , Because of this performance , Buy an extra computer , It's all done N Back and forth .
  • yes： start-up
  • no： Do not start （ Don't want to consume CPU resources , Can be turned off ）
• rdbchecksum： After storing the snapshot , Whether to start CRC64 Algorithm for data validation ;
  • After opening , About increase 10% Left and right CPU Consume ;
  • If you want to get the most performance improvement , Can choose to close
• dbfilename： Snapshot backup file name
  dir： The directory where the snapshot backup file is saved , Default to current directory

advantage and Inferiority

• optimal ： Suitable for large-scale data recovery , The requirements for data integrity and consistent rows are not high ;
• inferior ： Back up at regular intervals , accident down fall , You lose all the changes of the last snapshot

2. AOF

Append Only File

• Log every write operation ;
• take redis All write instructions executed are recorded （ Reading operation does not record ）;
• Only additional documents are allowed , You cannot overwrite the file ;
• redis At the beginning of startup, the file will be read and executed from beginning to end , To rebuild the data ;

2.1 Turn on AOF

1. To avoid mistakes , It is best to redis.conf Back up the general configuration file , Then modify the content as follows ：

appendonly yes
appendfilename appendonly.aof

2. Restart redis, Start with a new profile

redis-server /usr/local/redis5.0.4/redis.conf

3. Connect redis, Add data , Delete library , sign out

4. View the current folder one more aof file , Look at the contents of the file , It's all preserved Write operation

   1. The last sentence in the file should be deleted , Otherwise, the data cannot be recovered
   2. Edit this file , Finally :wq! Enforcement

5. Just reconnect , Data recovery successful

2.2 coexistence ？ Who is the first ？

We see the redis.conf file ,AOF and RDB Both backup strategies can be turned on at the same time , How will the system choose ？

1. Try it , edit appendonly.aof, Fooling around , Save and exit
2. start-up redis Failure , So it is AOF Load first to recover the original data ！ because AOF Than RDB The integrity of data preservation is higher ！
3. Repair AOF file , Killing all is not in line with redis The code of the grammar specification

reids-check-aof --fix appendonly.aof

2.3 And AOF Related configuration

• appendonly： Turn on aof Pattern
• appendfilename：aof File name of , Better not change ！
• appendfsync： Follow up strategy
  • always： Every data change , It will be recorded to disk immediately , Poor performance , But the data integrity is good
  • everysec： default setting , Asynchronous operations , Record every second , If it goes down in a second , There will be data loss
  • no： Don't follow up
• no-appendfsync-on-rewrite： Whether to use Appendfsync Follow up strategy ; Use default no that will do , Data security .
  • AOF By means of document addition , The file will get bigger and bigger , To solve this problem , Added rewriting mechanism ,redis Will automatically record the last AOF File size , When AOF When the file size reaches the preset size ,redis Will start AOF File content compression , Keep only the smallest set of instructions that can recover data
• auto-aof-rewrite-percentage： If AOF The file size has exceeded the original size 100%, That's double , Just rewrite the compression
• auto-aof-rewrite-min-size： If AOF The file has exceeded 64mb, Just rewrite the compression

3 summary （ How to choose ？）

• RDB： For backup purposes only , Suggest 15 Just back up once a minute
• AOF：
  • In the worst case , Only no more than 2 Second data , Data integrity is relatively high , But the price is too high , Will bring lasting IO
  • The size of the hard disk is also high , Default 64mb Is too small , The enterprise level is at least 5G above ;
  • master/slave Is the choice of sina Weibo ！！

End of the flower

If fate is the worst screenwriter in the world . Then you have to fight for , Be the best actor in your life

Writing is not easy to , If you think , Welcome to praise the blogger 、 Collection 、 A wave of comments ~ Let bloggers have more motivation