[PostgreSQL] PG91 的 Replication + Hot Stanby 配

轉自：http://godbms.blogspot.com/2011/11/pg91-replication-hot-stanby.html

1. 初始化
$ initdb -D path/to/pg911-prim --encoding=UTF-8 --no-locale --pwprompt --auth=md5

2. 免密碼設置
(hostname:port:database:username:password)
$ echo localhost:9112:postgres:go:go > ~/.pgpass
$ echo localhost:9113:postgres:go:go >> ~/.pgpass
$ chmod 600 ~/.pgpass

3. 創建 replication 用戶名密碼
$ psql -p 9112 -d postgres
psql (9.1.1)
Type "help" for help.

postgres=# CREATE ROLE reprole REPLICATION PASSWORD 'reppass';
CREATE ROLE
postgres=# ALTER ROLE reprole LOGIN;
ALTER ROLE

4. 修改 primary 配置
$ vim path/to/pg911-prim/postgresql.conf
listen_addresses = '*'
port = 9112
wal_level = hot_standby
max_wal_senders = 2 # standby + 1
wal_keep_segments = 8 # 8 - 32
synchronous_standby_names = ''
hot_standby = on # primary 中無效

$ vim path/to/pg911-prim/pg_hba.conf (追加)
host  replication  reprole  127.0.0.1/32  md5

5. 重啓動 primary, 使設置生效
$ pg_ctl restart -w

6. basebackup 到 stanby 目錄
$ pg_basebackup -x -h localhost -p 9112 -U reprole -D path/to/pg911-stby

7. 修改 stanby 配置
$ vim path/to/pg911-stby/postgres.conf
port = 9113

$ vim path/to/pg911-stby/recovery.conf
standby_mode = 'on'
primary_conninfo = 'host=localhost port=9112 user=reprole password=reppass application_name=stby'

8. 啓動 stanby
$ pg_ctl start -w
waiting for server to start....LOG:  database system was interrupted; last known up at 2011-11-16 15:36:32 JST
LOG:  creating missing WAL directory "pg_xlog/archive_status"
LOG:  entering standby mode
LOG:  redo starts at 0/2000020
LOG:  consistent recovery state reached at 0/3000000
LOG:  database system is ready to accept read only connections
LOG:  streaming replication successfully connected to primary
 done
server started

9. 查看啓動狀態
$ netstat -ano | grep PGSQL
unix  2      [ ACC ]     STREAM     LISTENING     50435  /tmp/.s.PGSQL.9113
unix  2      [ ACC ]     STREAM     LISTENING     49068  /tmp/.s.PGSQL.9112

10. 測試
$ psql -p 9112 -c "CREATE TABLE members(id integer, name text)" -d postgres
$ psql -p 9112 -c "INSERT INTO members VALUES(123, 'Foo')" -d postgres
$ psql -p 9112 -c "SELECT * FROM members" -d postgres
$ psql -p 9113 -c "SELECT * FROM members" -d postgres
$ psql -p 9113 -c "INSERT INTO members VALUES(456, 'Bar')" -d postgres

11. 提升 stanby 爲 primary (一旦升級後, 再降級要重新配置)
pg_ctl -D path/to/pg911-stby promote

12. 相關視圖
12-1 顯示 replication 狀態 (各參數說明？？？)
$ psql -x -p 9112 -c "SELECT * FROM pg_stat_replication" -d postgres
-[ RECORD 1 ]----+------------------------------
procpid          | 22856
usesysid         | 16384
usename          | reprole
application_name | stby
client_addr      | 127.0.0.1
client_hostname  |
client_port      | 34689
backend_start    | 2011-11-16 15:43:42.736069+09
state            | streaming
sent_location    | 0/301C028
write_location   | 0/301C028
flush_location   | 0/301C028
replay_location  | 0/301C028
sync_priority    | 0
sync_state       | async

12-2 primary 現在 wal 位置
$ psql -x -p 9112 -c "SELECT * FROM pg_current_xlog_location()" -d postgres

12-3 stanby 現在寫入 wal 位置
$ psql -x -p 9113 -c "SELECT * FROM pg_last_xlog_receive_location()" -d postgres

12-4 standby 到顯示的時間爲止的更新被應用
$ psql -p 9113 -c "SELECT pg_last_xact_replay_timestamp()"
 pg_last_xact_replay_timestamp
-------------------------------
 2011-04-25 23:20:56.912261+09
(1 row)

12-5 確認各進程
$ ps x | grep postgres
17829 pts/1    S      0:00 /usr/local/pgsql9.1.1/bin/postgres
17835 ?        Ss     0:00 postgres: writer process          
17836 ?        Ss     0:00 postgres: wal writer process      
17837 ?        Ss     0:00 postgres: autovacuum launcher process   
17838 ?        Ss     0:00 postgres: stats collector process   
17918 pts/1    S      0:00 /usr/local/pgsql9.1.1/bin/postgres
17923 ?        Ss     0:00 postgres: startup process   recovering 000000010000000000000004
17924 ?        Ss     0:00 postgres: writer process          
17925 ?        Ss     0:00 postgres: stats collector process   
17926 ?        Ss     0:00 postgres: wal receiver process   streaming 0/4088390
17927 ?        Ss     0:00 postgres: wal sender process reprole 127.0.0.1(26093) streaming 0/4088390
...

13 同期 replication 配置
13-1 修改配置
$ vim path/to/pg911-prim/postgresql.conf
synchronous_standby_names = 'stby' # 參照 path/to/pg911-stby/recovery.conf

13-2 重啓 primary 和 stanby
$ pg_ctl -D path/to/pg911-prim restart -w
$ pg_ctl -D path/to/pg911-stby restart -w

此時, 如果 stanby 停機, primary 的 commit 會被暫停, stanby 恢復後 commit 將成功
通過 synchronous_commit = local (on) 參數, 可以使 primary 單獨 (協同) 工作
$ pg_ctl -D path/to/pg911-prim -o "-c synchronous_commit=local" restart
$ pg_ctl -D path/to/pg911-prim -o "-c synchronous_commit=on" restart

14 primary/stanby 切換 (同期 or 非同期)
$ pg_ctl -D path/to/pg911-prim -m immediate stop
$ pg_ctl -D path/to/pg911-stby promote (新 primary)
$ vim path/to/pg911-prim/postgres.conf (修改)
hot_standby = on

$ vim path/to/pg911-prim/recovery.conf (創建)
standby_mode = 'on'
primary_conninfo = 'host=localhost port=9113 user=reprole password=reppass application_name=prim'
recovery_target_timeline = 'latest'
restore_command = 'cp /home/go/data/pg911-stby/pg_xlog/%f "%p" 2> /dev/null'

$ pg_ctl -D path/to/pg911-prim start -w (新 stanby)