Redis的持久化

edis支援RDB和AOF兩種持久化方式，持久化功能有效避免了因程序退出造成資料丢失的問題。

RDB持久化是把目前程序資料生成快照儲存到磁盤的過程，觸發RDB持久化的過程分為手動和自動觸發。

手動觸發RDB持久化的操作

1. 手動執行SAVE指令

執行指令：

127.0.0.1:6379> SAVE

OK

日志：

35650:M 18 Jul 18:10:42.795 * DB saved on disk

2. 手動執行BGSAVE指令

127.0.0.1:6379> BGSAVE

Background saving started

35650:M 18 Jul 18:12:48.613 * Background saving started by pid 39301

39301:C 18 Jul 18:12:48.803 * DB saved on disk

39301:C 18 Jul 18:12:48.805 * RDB: 8 MB of memory used by copy-on-write

35650:M 18 Jul 18:12:48.881 * Background saving terminated with success

BGSAVE是針對SAVE阻塞做的優化，Redis涉及RDB持久化的操作優先使用BGSAVE。

自動觸發RDB持久化的機制

1. save相關配置，如save m<seconds> n<changes>，m秒内資料集有n次修改，觸發BGSAVE。

配置：

save 900 1

save 300 10

save 60 10000

89428:M 18 Jul 18:46:49.809 * 10000 changes in 60 seconds. Saving...

89428:M 18 Jul 18:46:49.810 * Background saving started by pid 99472

99472:C 18 Jul 18:46:50.030 * DB saved on disk

99472:C 18 Jul 18:46:50.032 * RDB: 8 MB of memory used by copy-on-write

89428:M 18 Jul 18:46:50.113 * Background saving terminated with success

2. 建立主節點的一個從節點，主節點自動BGSAVE生成RDB檔案，發送給從節點。

從節點執行指令：

127.0.0.1:6389> SLAVEOF 127.0.0.1 6379

從節點日志：

99498:S 18 Jul 18:56:27.721 * SLAVE OF 127.0.0.1:6379 enabled (user request from 'id=2 addr=127.0.0.1:55238 fd=6 name= age=210 idle=0 flags=N db=0 sub=0 psub=0 multi=-1 qbuf=0 qbuf-free=32768 obl=0 oll=0 omem=0 events=r cmd=slaveof')

99498:S 18 Jul 18:56:27.829 * Connecting to MASTER 127.0.0.1:6379

99498:S 18 Jul 18:56:27.829 * MASTER <-> SLAVE sync started

99498:S 18 Jul 18:56:27.830 * Non blocking connect for SYNC fired the event.

99498:S 18 Jul 18:56:27.830 * Master replied to PING, replication can continue...

99498:S 18 Jul 18:56:27.830 * Partial resynchronization not possible (no cached master)

99498:S 18 Jul 18:56:27.911 * Full resync from master: 2f949fed88eb2df127e45a4870dce9fe1b37f2b2:1

99498:S 18 Jul 18:56:28.184 * MASTER <-> SLAVE sync: receiving 5241839 bytes from master

99498:S 18 Jul 18:56:28.330 * MASTER <-> SLAVE sync: Flushing old data

99498:S 18 Jul 18:56:28.330 * MASTER <-> SLAVE sync: Loading DB in memory

99498:S 18 Jul 18:56:28.388 * MASTER <-> SLAVE sync: Finished with success

主節點日志：

89428:M 18 Jul 18:56:27.831 * Slave 127.0.0.1:6389 asks for synchronization

89428:M 18 Jul 18:56:27.831 * Full resync requested by slave 127.0.0.1:6389

89428:M 18 Jul 18:56:27.831 * Starting BGSAVE for SYNC with target: disk

89428:M 18 Jul 18:56:27.907 * Background saving started by pid 99508

99508:C 18 Jul 18:56:28.105 * DB saved on disk

99508:C 18 Jul 18:56:28.106 * RDB: 8 MB of memory used by copy-on-write

89428:M 18 Jul 18:56:28.182 * Background saving terminated with success

89428:M 18 Jul 18:56:28.270 * Synchronization with slave 127.0.0.1:6389 succeeded

3. 執行DEBUG RELOAD重新加載Redis時，會觸發SAVE操作。

127.0.0.1:6379> DEBUG RELOAD

35650:M 18 Jul 18:40:22.379 * DB saved on disk

35650:M 18 Jul 18:40:22.433 # DB reloaded by DEBUG RELOAD

4. 執行SHUTDOWN時，會觸發SAVE操作。

127.0.0.1:6379> SHUTDOWN

89428:M 18 Jul 19:08:17.003 # User requested shutdown...

89428:M 18 Jul 19:08:17.003 * Saving the final RDB snapshot before exiting.

89428:M 18 Jul 19:08:17.098 * DB saved on disk

89428:M 18 Jul 19:08:17.098 * Removing the pid file.

89428:M 18 Jul 19:08:17.098 * Removing the unix socket file.

89428:M 18 Jul 19:08:17.098 # Redis is now ready to exit, bye bye...

RDB配置參數總覽：

stop-writes-on-bgsave-error yes

rdbcompression yes

rdbchecksum yes

dbfilename dump.rdb

AOF持久化，以獨立日志的方式記錄每次寫指令，解決了資料持久化的實時性問題。AOF的工作流程包括：append（指令寫入），sync（資料同步），rewrite（檔案重寫）和load（重新開機加載）。

AOF的rewrite過程可以手動觸發，和自動觸發。

1. 手動觸發

127.0.0.1:6379> BGREWRITEAOF

Background append only file rewriting started

99615:M 18 Jul 19:44:28.610 * Background append only file rewriting started by pid 99639

99615:M 18 Jul 19:44:28.900 * AOF rewrite child asks to stop sending diffs.

99639:C 18 Jul 19:44:28.900 * Parent agreed to stop sending diffs. Finalizing AOF...

99639:C 18 Jul 19:44:28.900 * Concatenating 0.00 MB of AOF diff received from parent.

99639:C 18 Jul 19:44:28.902 * SYNC append only file rewrite performed

99639:C 18 Jul 19:44:28.903 * AOF rewrite: 4 MB of memory used by copy-on-write

99615:M 18 Jul 19:44:28.923 * Background AOF rewrite terminated with success

99615:M 18 Jul 19:44:28.923 * Residual parent diff successfully flushed to the rewritten AOF (0.00 MB)

99615:M 18 Jul 19:44:28.923 * Background AOF rewrite finished successfully

2. 自動觸發

根據下面2個參數自動觸發。

auto-aof-rewrite-min-size 64mb，運作AOF重寫時，檔案最小空間。

auto-aof-rewrite-percentage 100，目前AOF檔案空間，和上次重寫後AOF檔案空間的比值。

AOF配置參數總覽：

appendonly yes

appendfilename "appendonly.aof"

appendfsync everysec

no-appendfsync-on-rewrite no

auto-aof-rewrite-percentage 100

auto-aof-rewrite-min-size 64mb

RDB和AOF兩種持久化方式對比：

1. RDB生成的是二進制檔案，占用空間小，恢複效率高。AOF生成的是遵守Redis序列化協定的文本檔案，占用空間大，恢複效率低。

~/keepmoving/6379/data $ file dump.rdb appendonly.aof 

dump.rdb:       data

appendonly.aof: ASCII text, with CRLF line terminators

2. RDB過程對作業系統造成的負載，一般比AOF要大，RDB是對全量資料集做快照，AOF是将資料集增量追加進日志檔案，其解決了資料持久化的實時性。