Shulou(Shulou.com)06/01 Report--

The usual design approach is:

The replication mechanism is used to compensate for the shortcomings of aof and snapshot performance, and the data can be persisted.

That is, RDB and AOF on master are not enabled to ensure master's read and write performance.

On the slave, RDB and AOF are enabled at the same time for persistence to ensure data security.

If the data is to be persisted and you want to ensure stability, it is recommended to leave half of the physical memory.

Because during snapshot, the child process that forks out to dump consumes the same memory as the parent process,

True copy-on-write, the impact on performance and memory consumption are relatively large.

Environment Description:

master: 192.168.2.100 Do not open RDB and AOF

slave: 192.168.2.200 Open RDB and AOF

Configuration information:

master：

# vim etc/redis.conf

#save 600 5 //disable RDB

appendonly no //disable AOF

requirepass 123456 //Specify authentication password

slave：

# vim etc/redis.conf

save 600 5 //Open RDB

dbfilename "dump_6379.rdb" //RDB file

dir "/usr/local/redis-3.0.6-6379" //RDB file path

appendonly yes //Open AOF

appendfilename "appendonly.aof" //Specify an AOF file

appendfsync everysec //Force writes to disk once per second

no-appendfsync-on-rewrite no //command appends are not performed during log rewriting

auto-aof-rewrite-percentage 100 //overwrite when current AOF exceeds previous AOF size by 100%

auto-aof-rewrite-min-size 64mb //log rewrite minimum

slaveof 192.168.2.100 6379 //Specify main library IP and port

masterauth 123456 //Specify the login password for the master library

Start redis:

master：# redis-server etc/redis.conf

slave：# redis-server etc/redis.conf

Master creates key:

# redis-cli -a 123456

127.0.0.1: 6379> info replication //Check whether the master-slave relationship is correct

127.0.0.1: 6379> keys * //At this time, there is no RDB file in the master installation directory.

(empty list or set)

127.0.0.1:6379> set name zhagnsan //Create 3 keys

OK

127.0.0.1:6379> set age 26

OK

127.0.0.1:6379> set home beijing

OK

slave check synchronization:

# redis-cli

127.0.0.1: 6379> keys * //data synchronized

1) "age"

2) "home"

3) "name"

Simulate master fault:

# ps -ef |grep redis

root 126472 1 0 21:58 ? 00:00:02 redis-server *:6379

root 127714 2844 0 22:29 pts/0 00:00:00 grep --color=auto redis

# kill -9 126472

# rm -rf dump_6379.rdb

slave View master-slave status:

# redis-cli

127.0.0.1:6379> info replication

# Replication

role:slave

master_host:192.168.2.100

master_port:6379

master_link_status:down //We see that master is no longer accessible, slave is normal

master_last_io_seconds_ago:-1

... ...

Disaster Recovery:

1. Cancel the synchronization status of slave

Avoid restarting the master without completing the data recovery, resulting in overwriting the data on the slave and eventually losing the data.

127.0.0.1: 6379> SLAVEOF no one //dynamically modify the configuration file to set slavof to no

OK

127.0.0.1: 6379> info repolarization//View master-slave information, no longer available

127.0.0.1:6379>

2. Start redis of master and confirm that the data does not exist (this step can be ignored, I just want to confirm that there is no data)

# redis-server etc/redis.conf

# redis-cli -a 123456

127.0.0.1:6379> keys *

(empty list or set)

# ps -ef |grep redis

root 128031 1 0 22:45 ? 00:00:00 redis-server *:6379

root 128050 2844 0 22:46 pts/0 00:00:00 grep --color=auto redis

# kill -9 128031 //abnormal close redis again

# rm -rf dump_6379.rdb //Delete RDB file again

3. Copy RDB files and AOF files to master

# scp dump_6379.rdb 192.168.2.100:/usr/local/redis-3.0.6-6379/

# scp appendonly.aof 192.168.2.100:/usr/local/redis-3.0.6-6379/

4. Master opens RDB, opens AOF

# vim etc/redis.conf

save 600 5 //Open RDB

dbfilename "dump_6379.rdb" //RDB file

dir "/usr/local/redis-3.0.6-6379" //RDB file path

appendonly yes //Open AOF

appendfilename "appendonly.aof" //Specify an AOF file

... ...

5. Master starts redis and checks the database. The data has been recovered.

# redis-server etc/redis.conf

# redis-cli -a 123456

127.0.0.1:6379> keys *

1) "home"

2) "name"

3) "age"

6. After master data recovery, RDB and AOF need to be turned off to ensure their own reading and writing performance.

However, RDB files and AOF files should be retained (deleted after data cannot be recovered)

Although RDB files and AOF files exist, they do not increase in size, only the AOF files of salve increase.

# kill [redis PID] //close redis

# vim etc/redis.conf //Close RDB and AOF

#save 600 5

appendonly no

# redis-server etc/redis.conf //start redis

# redis-cli -a 123456

127.0.0.1: 6379> keys * //Turn off RDB and AOF functions, the data in the library still exists

1) "home"

2) "name"

3) "age"

7. Slave opens synchronization state

127.0.0.1:6379> SLAVEOF 192.168.2.100 6379 //Start synchronization

OK

127.0.0.1:6379> info replication

# Replication

role:slave

master_host:192.168.2.100

master_port:6379

master_link_status:up //master can be accessed normally

master_last_io_seconds_ago:5

master_sync_in_progress:0

... ...

8. Master creates key

127.0.0.1: 6379> set abc 123 //Add a new key

OK

127.0.0.1:6379> keys *

1) "abc"

2) "home"

3) "name"

4) "age"

9. slave View synchronization

127.0.0.1:6379> keys *

1) "abc"

2) "age"

3) "home"

4) "name"

Summary:

During this recovery, we used both AOF and RDB files, so which file did the recovery take place?

1. If only RDB is configured, the RDB file is loaded at startup;

2. If only AOF is configured, the AOF file is loaded at startup;

If RDB and AOF are configured at the same time, AOF files are used because AOF has higher priority than RDB.

Furthermore, AOF files have higher data integrity than RDB files because RDBs are maintained according to a periodic policy.

If a failure occurs before the next cycle, data from the previous cycle to the point of failure will be lost.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.