PostgreSQL數據庫單機擴展爲流複製

 

 

1. 在standby服務器安裝postgres數據庫,不須要初始化.

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

2. 在primary服務器建立具備REPLICATION權限的複製用戶

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

3. 容許複製用戶遠程鏈接到primary服務器

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

4. 在primary服務器設置流複製相關的參數

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

5. 從新啓動primary服務器進程

1. $ pg_ctl stop -m fast 
2. $ pg_ctl start 

6. 對primary服務器作一個全備並傳輸到standby服務器

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

1. postgres=# SELECT pg_start_backup(‘label‘, true); 
2.  pg_start_backup  
3. ----------------- 
4.  7/E6000060 
5. (1 row) 
6. $ rsync -az --progress ${PGDATA} postgres@10.189.100.195:/usr/local/pgsql/ --exclude postmaster.pid 
7. postgres=# SELECT pg_stop_backup(); 
8. NOTICE:  pg_stop_backup complete, all required WAL segments have been archived 
9.  pg_stop_backup  
10. ---------------- 
11.  7/E60005C8 
12. (1 row) 

在standby服務器經過pg_basebackup命令進行備份,要求standby的PGDATA目錄爲空

1. $ 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 
2. Password:  
3. transaction log start point: 7/EA000028 on timeline 1 
4. pg_basebackup: starting background WAL receiver 
5. 65933562/65933562 kB (100%), 1/1 tablespace                                          
6. transaction log end point: 7/EA000830 
7. pg_basebackup: waiting for background process to finish streaming ... 
8. pg_basebackup: base backup completed 

7. 設置standby數據庫複製相關參數,使得standby失效轉移後能夠做爲主庫工做

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

8. 在standby文件建立恢復文件

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

9. 啓動standby數據庫進程,自動啓動流複製

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

10. 檢查primary和standby數據庫的延遲

• 經過函數和系統表查看

1. edbstore=# select * from pg_stat_replication;           #在primary主庫查看 
2. -[ RECORD 1 ]----+------------------------------ 
3. pid              | 15013 
4. usesysid         | 19206 
5. usename          | repl 
6. application_name | walreceiver 
7. client_addr      | 10.189.100.195 
8. client_hostname  |  
9. client_port      | 56072 
10. backend_start    | 2017-06-13 08:10:35.400508-07 
11. backend_xmin     |  
12. state            | streaming 
13. sent_location    | 7/EC01A588 
14. write_location   | 7/EC01A588 
15. flush_location   | 7/EC01A588 
16. replay_location  | 7/EC01A588 
17. sync_priority    | 0 
18. sync_state       | async 
19.  
20. edbstore=# SELECT pg_current_xlog_location();                      #在primary主庫查看 
21.  pg_current_xlog_location  
22. -------------------------- 
23.  7/EC01A588 
24. (1 row) 
25.  
26. postgres=# select pg_last_xlog_receive_location(),pg_last_xlog_replay_location(),pg_last_xact_replay_timestamp();     #在standby備庫查看 
27.  pg_last_xlog_receive_location | pg_last_xlog_replay_location | pg_last_xact_replay_timestamp  
28. -------------------------------+------------------------------+------------------------------- 
29.  7/EC01A588                    | 7/EC01A588                   | 2017-06-13 08:25:20.281568-07 
30. (1 row) 

• 經過進程查看

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