PostgreSQL 基於日志的備份與還原

參考：http://www.postgresql.org/docs/9.5/static/continuous-archiving.html

http://www.mkyong.com/database/postgresql-point-in-time-recovery-incremental-backup/

 

wal，即預寫式日志，是日志的標准實現方式，簡單而言就是將對數據庫的變動記錄到日志 中，而后在將具體的新數據刷新到磁盤。PostgreSQL將該日志維護在數據文件夾下的子文件夾pg_xlog中。當數據庫崩潰后，可以通過“重放”日志中的“動作”，將數據庫恢復。也就是說，只要擁有一個基礎備份和完整的日志文件，理論上可以將數據庫庫恢復到任意基礎備份以來的任意時刻點。不僅如此，如果在另一個實例上將這些日志不停的“重放”，那么就擁有了一個完整的在線備份，也就是“復制”。

pg_xlog下日志文件不會無限制增多，也就是說並不用擔心日志的增多使得磁盤空間捉襟見肘。默認每個日志文件為16M大小，即當增長到16M時，就會切換到別的文件，並復用之前的文件 。因此，為了保證有個完整的日志鏈，必須將寫滿的文件復制保存到一個特定的文件 夾。對於最后一個還未滿16M的日志文件，可以手動觸發一次切換。

 

備份（操作均使用系統賬號postgres完成 ）

1. 登錄數據庫創建測試數據庫

postgres@debian:~$ psql
psql (9.5.0)

Type "help" for help.

postgres=# CREATE DATABASE test;

CREATE DATABASE

 

2. 修改配置文件，開啟日志備份，將寫滿的文件復制到archive文件夾下

vim /etc/postgresql/9.5/main/postgresql.conf
wal_level = archive

archive_mode = on

archive_command = ' test ! -f /var/lib/postgresql/archive/%f && cp %p /var/lib/postgresql/archive/%f'

 

3. 創建archive文件夾，並重啟數據庫服務

postgres@debian:~$ mkdir archive

postgres@debian:~$ /usr/lib/postgresql/9.5/bin/pg_ctl restart -D /var/lib/postgresql/9.5/main/ -o "-c config_file=/etc/postgresql/9.5/main/postgresql.conf"

2016-01-18 09:30:42 CST [2937-2] LOG:  received fast shutdown request

2016-01-18 09:30:42 CST [2937-3] LOG:  aborting any active transactions

2016-01-18 09:30:42 CST [2942-2] LOG:  autovacuum launcher shutting down

2016-01-18 09:30:42 CST [2939-1] LOG:  shutting down

waiting for server to shut down.....2016-01-18 09:30:44 CST [2939-2] LOG:  database system is shut down

 done

server stopped

server starting

postgres@debian:~$ 2016-01-18 09:30:45 CST [2972-1] LOG:  database system was shut down at 2016-01-18 09:30:44 CST

2016-01-18 09:30:45 CST [2972-2] LOG:  MultiXact member wraparound protections are now enabled

2016-01-18 09:30:45 CST [2971-1] LOG:  database system is ready to accept connections

2016-01-18 09:30:45 CST [2976-1] LOG:  autovacuum launcher started

4. 創建測試表

postgres@debian:~$ psql

psql (9.5.0)

Type "help" for help.

postgres=# \c test

You are now connected to database "test" as user "postgres".

test=# CREATE TABLE testPITR1 AS SELECT * FROM pg_class, pg_description;

SELECT 1192063

5. 創建基礎備份

psql -c "SELECT pg_start_backup('base', true)"

cd /var/lib/postgresql/9.5/

tar -cvf main.tar main

psql -c "SELECT pg_stop_backup()"

　 6. 繼續創建測試表，切換日志

postgres@debian:~$ psql

psql (9.5.0)

Type "help" for help.

 

postgres=# \c test

You are now connected to database "test" as user "postgres".

test=# CREATE TABLE testPITR2 AS SELECT * FROM pg_class, pg_description;

SELECT 1203562

test=#  select * from current_timestamp;

              now             

-------------------------------

 2016-01-18 10:02:15.229335+08

(1 row)

 

test=# CREATE TABLE testPITR3 AS SELECT * FROM pg_class, pg_description;

SELECT 1215061

test=#  select * from current_timestamp;

              now             

-------------------------------

 2016-01-18 10:02:51.029447+08

(1 row)

test=# select pg_switch_xlog();

 pg_switch_xlog

----------------

 0/3DDE6750

(1 row)

恢復

關閉數據庫，模擬數據庫宕機，此時，數據庫test中應該有3張表，其中1張表在基礎備份前，也就是恢復完數據文件即可找回，而另2張表則需恢復相應的日志文件。模擬恢復到testPITR2創建時刻點。

1. 關閉數據庫服務，重命名數據文件夾

postgres@debian:~$ /usr/lib/postgresql/9.5/bin/pg_ctl stop -D /var/lib/postgresql/9.5/main/

2016-01-18 10:06:12 CST [2971-2] LOG:  received fast shutdown request

2016-01-18 10:06:12 CST [2971-3] LOG:  aborting any active transactions

2016-01-18 10:06:12 CST [2976-2] LOG:  autovacuum launcher shutting down

2016-01-18 10:06:12 CST [2973-1] LOG:  shutting down

waiting for server to shut down.....2016-01-18 10:06:13 CST [2973-2] LOG:  database system is shut down

 done

server stopped

postgres@debian:~$ mv 9.5/main 9.5/main.old

2. 解壓備份數據文件，開啟服務，驗證此時只有基礎備份前的表testpitr1

postgres@debian:~$cd /var/lib/postgresql/9.5/

postgres@debian:~/9.5$ tar -xvf 9.5/main.tar

postgres@debian:~$ 2016-01-18 10:26:40 CST [3342-1] LOG:  database system was interrupted; last known up at 2016-01-18 09:54:56 CST

2016-01-18 10:26:40 CST [3342-2] LOG:  redo starts at 0/17000098

2016-01-18 10:26:40 CST [3342-3] LOG:  invalid record length at 0/17009348

2016-01-18 10:26:40 CST [3342-4] LOG:  redo done at 0/170092D8

2016-01-18 10:26:40 CST [3342-5] LOG:  last completed transaction was at log time 2016-01-18 09:48:26.585085+08

2016-01-18 10:26:40 CST [3342-6] LOG:  MultiXact member wraparound protections are now enabled

2016-01-18 10:26:40 CST [3341-1] LOG:  database system is ready to accept connections

2016-01-18 10:26:40 CST [3346-1] LOG:  autovacuum launcher started

 

postgres@debian:~$ psql

psql (9.5.0)

Type "help" for help.

 

postgres=# \c test

You are now connected to database "test" as user "postgres".

test=# \d

           List of relations

 Schema |   Name    | Type  |  Owner  

--------+-----------+-------+----------

 public | testpitr1 | table | postgres

(1 row)

3. 停掉數據庫，刪除數據文件夾，重新解壓基礎備份，創建recovery.conf，將數據庫恢復到testpitr2時刻

vi recovery.conf

restore_command = 'cp /var/lib/postgresql/archive/%f %p'

recovery_target_time = '2016-01-18 10:02:15'

postgres@debian:~/9.5/main$ /usr/lib/postgresql/9.5/bin/pg_ctl start -D /var/lib/postgresql/9.5/main/ -o "-c config_file=/etc/postgresql/9.5/main/postgresql.conf" -l /var/lib/postgresql/recovery.log

可以在恢復日志中看到這么一句話：

2016-01-18 11:22:39 CST [1743-44] LOG:  recovery stopping before commit of transaction 630, time 2016-01-18 10:02:46.080443+08

4. 驗證，進入數據庫，發現testpitr2已經恢復

postgres@debian:~$ psql

psql (9.5.0)

Type "help" for help.

 

postgres=# \c test

You are now connected to database "test" as user "postgres".

test=# \dt

           List of relations

 Schema |   Name    | Type  |  Owner  

--------+-----------+-------+----------

 public | testpitr1 | table | postgres

 public | testpitr2 | table | postgres

(2 rows)

 

如果需要恢復table3，則必須再次刪除數據文件夾，建立recovery.conf。