[[194921]]

1. 在standby服務器安裝postgres數據庫,不需要初始化.

安裝過程詳見:http://www.cnblogs.com/ilifeilong/p/6979288.html

2. 在primary服務器創建具有REPLICATION權限的復制用戶

postgres=# CREATE ROLE repl WITH REPLICATION PASSWORD ‘repl‘ LOGIN; 

3. 允許復制用戶遠程連接到primary服務器

$ grep "^host" pg_hba.conf 
host    all             all             127.0.0.1/32            trust 
host    replication             repl             0.0.0.0/0               md5  
host    all             all             ::1/128                 trust 

4. 在primary服務器設置流復制相關的參數

$ mkdir /usr/local/pgsql/arch  
$ egrep "archive_mode|max_wal_senders|wal_keep_segments|archive_command|wal_level|hot_standby" postgresql.conf 
al_level = hot_standby            # minimal, archive, hot_standby, or logical 
archive_mode = on        # enables archiving; off, on, or always 
archive_command = ‘test ! -f /usr/local/pgsql/arch/%f && cp %p /usr/local/pgsql/arch/%f‘         
max_wal_senders = 5        # max number of walsender processes 
wal_keep_segments = 30        # in logfile segments, 16MB each; 0 disables 
hot_standby = on            # "on" allows queries during recovery 
#hot_standby_feedback = off        # send info from standby to prevent 

5. 重新啟動primary服務器進程

$ pg_ctl stop -m fast 
$ pg_ctl start 

6. 對primary服務器做一個全備并傳輸到standby服務器

• 在primary服務器通過pg_(start|stop)_backup函數進行備份

postgres=# SELECT pg_start_backup(‘label‘, true); 
 pg_start_backup  
----------------- 
 7/E6000060 
(1 row) 
$ rsync -az --progress ${PGDATA} postgres@10.189.100.195:/usr/local/pgsql/ --exclude postmaster.pid 
postgres=# SELECT pg_stop_backup(); 
NOTICE:  pg_stop_backup complete, all required WAL segments have been archived 
 pg_stop_backup  
---------------- 
 7/E60005C8 
(1 row) 

在standby服務器通過pg_basebackup命令進行備份,要求standby的PGDATA目錄為空

$ pg_basebackup --host=10.189.102.118 --username=repl --port=5432 --label=backup --verbose --progress --pgdata=/usr/local/pgsql/data --checkpoint=fast --format=p --xlog-method=stream 
Password:  
transaction log start point: 7/EA000028 on timeline 1 
pg_basebackup: starting background WAL receiver 
65933562/65933562 kB (100%), 1/1 tablespace                                          
transaction log end point: 7/EA000830 
pg_basebackup: waiting for background process to finish streaming ... 
pg_basebackup: base backup completed 

7. 設置standby數據庫復制相關參數,使得standby失效轉移后可以作為主庫工作

$ mkdir /usr/local/pgsql/arch 
$ egrep "archive_mode|max_wal_senders|wal_keep_segments|archive_command|wal_level|hot_standby" postgresql.conf 
wal_level = hot_standby                 # minimal, archive, hot_standby, or logical 
archive_mode = on               # enables archiving; off, on, or always 
archive_command = ‘test ! -f /usr/local/pgsql/arch/%f && cp %p /usr/local/pgsql/arch/%f‘ 
max_wal_senders = 5             # max number of walsender processes 
wal_keep_segments = 30          # in logfile segments, 16MB each; 0 disables 
hot_standby = on                        # "on" allows queries during recovery 
#hot_standby_feedback = off             # send info from standby to prevent 

8. 在standby文件創建恢復文件

$ cat recovery.conf  
restore_command = ‘cp /usr/local/pgsql/arch/%f "%p"‘ 
standby_mode = ‘on‘ 
primary_conninfo = ‘user=repl password=repl host=10.189.102.118 port=5432 sslmode=disable sslcompression=1‘ 
archive_cleanup_command = ‘pg_archivecleanup -d /usr/local/pgsql/arch %r >> /usr/local/pgsql/arch/archive_cleanup.log‘ 
trigger_file = ‘/usr/local/pgsql/data/trigger_active_standby‘ 

9. 啟動standby數據庫進程,自動啟動流復制

$ pg_ctl start -w 
waiting for server to start....LOG:  could not create IPv6 socket: Address family not supported by protocol 
LOG:  redirecting log output to logging collector process 
HINT:  Future log output will appear in directory "pg_log". 
 done 
server started 

10. 檢查primary和standby數據庫的延遲

• 通過函數和系統表查看

edbstore=# select * from pg_stat_replication;           #在primary主庫查看 
-[ RECORD 1 ]----+------------------------------ 
pid              | 15013 
usesysid         | 19206 
usename          | repl 
application_name | walreceiver 
client_addr      | 10.189.100.195 
client_hostname  |  
client_port      | 56072 
backend_start    | 2017-06-13 08:10:35.400508-07 
backend_xmin     |  
state            | streaming 
sent_location    | 7/EC01A588 
write_location   | 7/EC01A588 
flush_location   | 7/EC01A588 
replay_location  | 7/EC01A588 
sync_priority    | 0 
sync_state       | async 
 
edbstore=# SELECT pg_current_xlog_location();                      #在primary主庫查看 
 pg_current_xlog_location  
-------------------------- 
 7/EC01A588 
(1 row) 
 
postgres=# select pg_last_xlog_receive_location(),pg_last_xlog_replay_location(),pg_last_xact_replay_timestamp();     #在standby備庫查看 
 pg_last_xlog_receive_location | pg_last_xlog_replay_location | pg_last_xact_replay_timestamp  
-------------------------------+------------------------------+------------------------------- 
 7/EC01A588                    | 7/EC01A588                   | 2017-06-13 08:25:20.281568-07 
(1 row) 

• 通過進程查看

$ ps -ef | grep sender | grep -v grep #在primary庫查看  
postgres 15013 24883 0 08:10 ? 00:00:00 postgres: wal sender process repl 10.189.100.195(56072) streaming 7/EC01A668  
$ ps -ef | grep receiver | grep -v grep #在standby庫查看  
postgres 12857 12843 0 08:10 ? 00:00:00 postgres: wal receiver process streaming 7/EC01A668