Redis Persistence - RDB and AOF

Redis 的数据是存储在内存当中, because the server is down、停电等情况, Redis The data in it is easy to lose.这时 Redis The role of persistence is reflected.例: Can be used to complete data recovery、failover data, etc.

Redis Persistence is divided into three policy mechanisms: RDB（Redis DataBase）、 AOF（Append Only File）、Mixed persistence in two ways.

This article briefly describes both methods. For in-depth knowledge points, Congratulations, everyone, please enter the advanced area to check.

点我查看 - Redis 基本知识,快来回顾一下

点我查看 - Redis 订阅与 Redis Stream 技术

点我查看 - PHP 使用 PHPRedis 与 Predis

一、RDB（Redis DataBase）

RDB 是指周期性Writes an in-memory snapshot of the dataset to a disk file.

In this way, the data in memory is written to the binary file in the form of a snapshot, 默认的文件名为 dump.rdb, 在 Redis 重启时, 通过加载 dump.rdb 文件来恢复数据.

---

1.1 RDB 的触发方式

1.1.1 自动触发

顾名思义, 自动触发是 Redis 在某个时间内, Automate persistence operations, 将数据进行备份.

Redis 配置文件是 /usr/local/redis/bin/redis.conf.

• 这是我的 redis.conf 所在位置, Find the file according to your own situation

• 打开后, 我的在 412 The line starts with the configuration for the snapshot

• 我这里默认 RDB 持久化功能是关闭的, # save "", 我的在 424 行

• dbfilename dump.rdb 文件名配置, 我的在 481 行

• dir ./ The backup file holds the directory configuration, 我的在 504 行

• By default it is annotated, 如果要启用, will start with the line # 去掉

# 配置格式: save m n [m n ...]
# 解释: 在 m 秒内, 如果有 n 个键发生改变, The persistence operation is automatically triggered, Multiple conditions can be configured

# 例: save 3600 1 300 100 60 10000
# 例解: 3600 秒内, 如果有 1 个键发生改变, The persistence operation is automatically triggered
# 例解: 300 秒内, 如果有 100 个键发生改变, The persistence operation is automatically triggered
# 例解: 60 秒内, 如果有 10000 个键发生改变, The persistence operation is automatically triggered

---

1.1.2 手动触发

顾名思义, Manual trigger is pass 手动执行命令 来完成持久化操作.

There are two commands that are triggered manually: save()、bgsave(), The main difference between them is whether or not they block the thread.

SAVE 命令:

• save() 命令触发 Redis 同步持久化.
• 缺点: 会使 Redis 处于阻塞状态, 不能执行其他命令, 直到 RDB 持久化完成.
• 优点: Doesn't consume extra memory though,但生产环境慎用.

BGSAVE 命令:

bgsave() 会执行 fork 操作创建子进程, The persistent work is done by the child process, 整个过程只有在 fork There is a brief block when the child process, When the child process is created, It can respond to requests from other clients.

• bgsave() 命令触发 Redis 异步持久化.
• 缺点: 创建子进程, 需要消耗更多内存.
• 优点: 创建子进程后, Other clients can also interact, Used in most casesbgsave().

---

1.2 RDB 的优缺点

1.2.1 RDB 优点

• RDB The persistent file is binary data, 占用内存更小.
• RDB Because the backup is a snapshot of the database, So it is very suitable for cold backup, 比如 凌晨1点的时候, 进行持久化操作.
• in the face of big data, RDB Backup to restore speed ratio AOF 快.

---

1.2.2 RDB 缺点

• RDB Backups are performed periodically, 如果发生意外情况, And there is data interaction during this time, Data for this period will be lost.
• 无法实现 实时持久化.
• Because it is a full backup, And also often create subprocesses, 数据量大的话, Will consume the server for a long timeCPU与内存.

---

二、AOF（Append Only File）

AOF Record in log, 将执行完的命令、命令参数、Information such as the number of parameters is passed write The function appends to the end of the log file.

2.1 同步流程

Command synchronization process：

• Send commands or parameters, etc. to AOF 程序
• AOF After the program is received, 因为 Redis 是单线程, 通过 resp 协议, Staging data to AOF 缓冲区
• By setting the synchronization policy, Synchronously appends commands from the staging area to the end of the file

---

2.2 同步策略

# Synchronization strategy by appendfsync 参数决定, 我的在 1435 行开始
# appendfsync always

appendfsync everysec # 默认

# appendfsync no

• always

  AOF When the buffer receives data, 立即调用 fsync The command is forced to go AOF 文件写数据.

• everysec

  AOF When the buffer receives data, 调用 write 命令往 io 缓冲区写数据, Also called once per second fsync 命令, 强制将 io 缓冲区的数据写入 AOF 文件.

• no

  AOF When the buffer receives data, 调用 write 命令往 io 缓冲区写数据, **当 io When the buffer fills up or periodically the system will io 缓冲区的数据写入 AOF 文件 **.

---

2.3 AOF 重写

As more and more commands are executed, AOF 文件会越来越大, 可采用 AOF 重写机制 解决这一问题.

Overriding enables a background child process bgrewriteaof.

2.3.1 配置

AOF 重写 是去除 AOF 文件中 无效的命令、超时的数据 wait or adopt 命令合并 方式, 创建一个新的 AOF file to hold the rewritten commands.

no-appendfsync-on-rewrite no # The rewrite mechanism is disabled by default, 如需开启, 请改为 yes , 我的在 1460 行

---

2.3.2 触发重写

• 手动触发

  AOF 重写可以由用户通过调用 BGREWRITEAOF 命令手动触发.

• 自动触发

  根据 auto-aof-rewrite-percentage 与 auto-aof-rewrite-min-size 64mb 确定触发时机.

  # 我的在 1479 行
  auto-aof-rewrite-percentage 100 # 当前 AOF 文件大小(aof_current_size) 比 The size after the previous rewrite(aof_base_size) 至少大了 100%
  
  auto-aof-rewrite-min-size 64mb # AOF 体积大于 64mb 会触发重写
  

2.4 开启 AOF

默认 AOF 持久化功能是关闭的.

Redis 配置文件是 /usr/local/redis/bin/redis.conf.打开后, 我的在 1359 line is the start of the configuration.

appendonly yes # 开启, 改为 yes , 我的在 1379 行

appendfilename "appendonly.aof" # Snapshot base filename

appenddirname "appendonlydir" # Storage dedicated directory

---

2.5 AOF 的优缺点

2.5.1 AOF 的优点

• Data integrity can be better protected, 比 RDB 更靠谱
• Even if the file is too large, 也可以采用 重写 方式, 不影响客户端

2.5.2 AOF 的缺点

• Because of the synchronization strategy, 一般情况下 AOF 的速度慢于 RDB
• Even if it's been rewritten all the time, But the file size is still better RDB The binary file is large
• AOF 在恢复时, Each command will be repeated, 速度也比 RDB 慢
• Due to its large size, 恢复时, There may be write corruption, You can compare the differences between the two files, 使用 redis-check-aof --fix AOF文件名 修复, diff -u to compare the differences in the files, And then confirm and then restore.

---

三、两者比较

方式	存储速度	恢复速度	数据完整性	文件大小	资源消耗
RDB	慢	快	不可抗拒因素, 可能会丢失数据	小	高
AOF	快	慢	According to the synchronization strategy,视情况而定, May be badly written	大	低

Choose the appropriate persistence method according to the business scenario, 默认 RDB .

• If the business scenario mainly acts as a cache function, Failed to revisit data acquisition, Less reliance on persistence, 则选择 RDB.
• 如果数据需要持久保存, 确保数据的完整性, 则建议 RDB 与 AOF 同时启用.

---

点我查看 - Redis 基本知识,快来回顾一下

点我查看 - Redis 订阅与 Redis Stream 技术

点我查看 - PHP 使用 PHPRedis 与 Predis