1. postgres 主備環境搭建

1.1  HA 實現方式

      熱備份、冷備份

     冷備份：物理備份，也就是檔案複製。主機寫完一個WAL日誌檔案後，才複製到standby伺服器上去應用。

     熱備份：流複製，也就是同步複製或非同步複製。主備機實時同步。

1.2 操作步驟

主備機安裝在相同的機器上

（1）主備機PGDATA路徑

primary PGDATA路徑：
[[email protected] data]$ echo $PGDATA
/home/wln/postgres9.3/data

standby PGDATA路徑：

[[email protected]

（2）修改primary 的postgresql.conf 檔案

wal_level = hot_standby   //熱備模式
max_wal_senders = 2    //primary 最多允許的併發standby資料庫
wal_keep_segments = 32  //in logfile segments, 16MB each; 0 disables。該xlog日誌會迴圈覆蓋，請將該值設定適當大些（需考慮磁碟空間、資料量增長情況）

logging_collector = on

（3）修改primary的pg_hba.conf

host replication repluser 127.0.0.1/32 trust

資料庫名必須填“replication”， 這是一個為standby連線使用了一個虛擬的資料庫名稱。使用者repluser是給standby連線使用的在主庫上建的一個超級使用者。
127.0.0.1 是standby資料庫的IP地址。

（4）建立一個數據庫超級使用者repluser,用來給standby備機連線主機使用。

[[email protected] data]$ psql -d postgres
psql (9.3beta2)
Type "help" for help.

postgres=# create user repluser superuser password 'password';
CREATE ROLE

（5）重啟primary資料庫

pg_ctl restart

（6）對主資料庫做一個基礎備份
用select pg_start_backup();命令把資料庫切換到備份狀態

[[email protected] data]$ psql -d postgres
psql (9.3beta2)
Type "help" for help.

postgres=# select pg_start_backup('/home/wln/standby');
 pg_start_backup
-----------------
 0/6000024
(1 row)

（不要退出）

（7）新開1個session2，將主機資料目錄copy到備機

[[email protected] ~]$ cp -r postgres9.3/data/* standby/
[[email protected] ~]$ cd standby/

[[email protected] standby]$ ll
總計 104
-rw------- 1 wln wln   205 07-06 04:13 backup_label
drwx------ 5 wln wln  4096 07-06 04:13 base
drwx------ 2 wln wln  4096 07-06 04:13 global
drwx------ 2 wln wln  4096 07-06 04:13 pg_clog
-rw------- 1 wln wln  4461 07-06 04:13 pg_hba.conf
-rw------- 1 wln wln  1636 07-06 04:13 pg_ident.conf
drwx------ 2 wln wln  4096 07-06 04:13 pg_log
drwx------ 4 wln wln  4096 07-06 04:13 pg_multixact
drwx------ 2 wln wln  4096 07-06 04:13 pg_notify
drwx------ 2 wln wln  4096 07-06 04:13 pg_serial
drwx------ 2 wln wln  4096 07-06 04:13 pg_snapshots
drwx------ 2 wln wln  4096 07-06 04:13 pg_stat
drwx------ 2 wln wln  4096 07-06 04:13 pg_stat_tmp
drwx------ 2 wln wln  4096 07-06 04:13 pg_subtrans
drwx------ 2 wln wln  4096 07-06 04:13 pg_tblspc
drwx------ 2 wln wln  4096 07-06 04:13 pg_twophase
-rw------- 1 wln wln     4 07-06 04:13 PG_VERSION
drwx------ 3 wln wln  4096 07-06 04:13 pg_xlog
-rw------- 1 wln wln 20312 07-06 04:13 postgresql.conf
-rw------- 1 wln wln    43 07-06 04:13 postmaster.opts
-rw------- 1 wln wln    83 07-06 04:13 postmaster.pid

（8）copy完成，結束主機備份狀態

postgres=# select pg_stop_backup();
NOTICE:  WAL archiving is not enabled; you must ensure that all required WAL segments are copied through other means to complete the backup
 pg_stop_backup
----------------
 0/60000DC
(1 row)

（9）在session2上操作

修改備機postgresql.conf檔案

port=5433

hot_standby=on

將share/recovery.conf.sample  複製到standby目錄下，並修改如下：

standby_mode = 'on'
primary_conninfo = 'host=127.0.0.1 port=5432 user=repluser password=password'
trigger_file = '/home/wln/standby/trigger_activestb'

刪除standby目錄下的postmaster.pid檔案，執行

[[email protected] standby]$export PGDATA=/home/wln/standby

[[email protected] standby]$ echo $PGDATA
/home/wln/standby

（10）啟動備機

在session2操作

pg_ctl start

檢查有沒有正常啟動pg_ctl status

若正常啟動，pg_log下日誌：

LOG:  database system was shut down at 2014-07-06 03:55:40 CST
LOG:  database system is ready to accept connections
LOG:  autovacuum launcher started
LOG:  database system was interrupted; last known up at 2014-07-06 04:13:09 CST
LOG:  entering standby mode
LOG:  redo starts at 0/6000024
LOG:  record with zero length at 0/60000B8
LOG:  started streaming WAL from primary at 0/6000000 on timeline 1
LOG:  consistent recovery state reached at 0/60000DC

（11）主機上執行一些操作，檢視備機有沒有同步過來

主機上：

postgres=# \d
No relations found.
postgres=# create table t1(id int);
CREATE TABLE
postgres=# insert into t1 select generate_series(1,10);
INSERT 0 10

postgres=# select pg_current_xlog_location();
 pg_current_xlog_location 
--------------------------
 0/7034470
(1 row)

postgres=# select * from pg_stat_replication ;
-[ RECORD 1 ]----+------------------------------
pid              | 10183
usesysid         | 24715
usename          | repl
application_name | walreceiver
client_addr      | 127.0.0.1
client_hostname  | 
client_port      | 41560
backend_start    | 2014-07-06 04:15:45.185796+08
state            | streaming
sent_location    | 0/7034470
write_location   | 0/7034470
flush_location   | 0/7034470
replay_location  | 0/7034470
sync_priority    | 0
sync_state       | async

備機上：

postgres=# \d
       List of relations
 Schema | Name | Type  | Owner
--------+------+-------+-------
 public | t1   | table | wln
(1 row)

postgres=# select * from t1;
 id
----
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
(10 rows)

postgres=# select pg_last_xlog_receive_location();
 pg_last_xlog_receive_location 
-------------------------------
 0/7034470
(1 row)

postgres=# select * from pg_stat_replication ;
(No rows)
postgres=# create table  t2(id int);
ERROR:  cannot execute CREATE TABLE in a read-only transaction

（12）一些引數設定

synchronous_standby_names = '*' # standby servers that provide sync rep
                                # comma-separated list of application_name
                                # from standby(s); '*' = all

（預設為‘’）

synchronous_commit = on                # synchronization level;
                                        # off, local, remote_write, or on

Commits made when synchronous_commit is set to on or remote_write will wait until the synchronous standby responds. The response may never occur if the last, or only, standby should crash.

The best solution for avoiding data loss is to ensure you don't lose your last remaining synchronous standby. This can be achieved by naming multiple potential synchronous standbys using synchronous_standby_names. The first named standby will be used as the synchronous standby. Standbys listed after this will take over the role of synchronous standby if the first one should fail.

3. 主備切換(方式1)

主備不會自動切換，需藉助第三方軟體，如heartbeat等，請自行搜尋。

（1）如何檢視是primary還是standby

方法1：檢視程序

主機 sender

[[email protected] data]$ ps -ef | grep wal
wln      10421 10416  0 07:52 ?        00:00:00 postgres: wal writer process             
wln      10691 10416  0 08:04 ?        00:00:00 postgres: wal sender process repluser 127.0.0.1(59226) streaming 0/6011BAC
wln      10702  8927  0 08:05 pts/2    00:00:00 grep wal

備機 receiver

[[email protected] data]$ ps -ef | grep wal
wln      10421 10416  0 07:52 ?        00:00:00 postgres: wal writer process             
wln      10690 10685  3 08:04 ?        00:00:03 postgres: wal receiver process   streaming 0/6011BAC
wln      10702  8927  0 08:05 pts/2    00:00:00 grep wal

方法2：

主機：

[[email protected] data]$ pg_controldata | grep Database
Database system identifier:           6032728846531166188
Database cluster state:               in production
Database block size:                  8192

 備機：

[[email protected] standby]$ pg_controldata | grep Database
Database system identifier:           6032728846531166188
Database cluster state:               in archive recovery
Database block size:                  8192

（2）備機切換為主機

將主機停掉

[[email protected] data]$ pg_ctl stop
waiting for server to shut down.... done
server stopped

在備機PGDATA對應目錄下生成recovery.conf檔案中指定的檔案trigger_file = '/home/wln/standby/trigger_activestb'  （此時備機正常執行狀態）

touch  /home/wln/standby/trigger_activestb

會看到 recovery.conf 檔案變為 recovery.done

檢視原來備機日誌：

LOG:  trigger file found: /home/wln/standby/trigger_activestb
LOG:  redo done at 0/6011C40
LOG:  last completed transaction was at log time 2014-07-06 08:05:40.152147+08
LOG:  selected new timeline ID: 2
LOG:  archive recovery complete
LOG:  database system is ready to accept connections
LOG:  autovacuum launcher started

登入原來的備機

[[email protected] standby]$ psql -d postgres -p 5433
psql (9.3beta2)
Type "help" for help.

postgres=# show transaction_read_only;
 transaction_read_only
-----------------------
 off
(1 row)

可以看出現在原備機具有讀寫功能。即主備切換成功

postgres=# \d
       List of relations
 Schema | Name | Type  | Owner
--------+------+-------+-------
 public | t1   | table | wln
(1 row)

postgres=# create table t2(id int);
CREATE TABLE

4、主備切換（方式2）

備機中recovery.conf中設定為：

standby_mode = on

recovery_target_timeline = 'latest'

primary_conninfo = 'host=localhost port=5433 user=repluser password=password'    

#trigger_file = '/home/wln/standby/trigger_activestb'  #這個註釋掉

 備機升為主機方式為：pg_ctl promote -D standbyPath

會看到recovery.conf過幾秒後變為recovery.done 表示升級為主機成功。

其他的請參考方式1中內容。

 5. 簡單指令碼

#master
mDATA=/home/$USER/pg94/data
port=5432
wal_level=hot_standby
max_wal_senders=2
wal_keep_segments=5
logging_collector=on
log_line_prefix='%d %p %t %c %i %e '
#standby
sDATA=/home/$USER/pg94/standby
mkdir $sDATA
chmod 700 $sDATA
sport=5433
shot_standby=on

pg_ctl stop -m i
pg_ctl stop -D $sDATA -m i
rm -rf $sDATA/*

#step
sed -i /port\ =/c\port\ =\ $port                                        $mDATA/postgresql.conf
sed -i /wal_level\ =/c\wal_level\ =\ $wal_level                         $mDATA/postgresql.conf
sed -i /max_wal_senders\ =/c\max_wal_senders\ =\ $max_wal_senders       $mDATA/postgresql.conf
sed -i /wal_keep_segments\ =/c\wal_keep_segments\ =\ $wal_keep_segments $mDATA/postgresql.conf
sed -i /logging_collector\ =/c\logging_collector\ =\ $logging_collector $mDATA/postgresql.conf
sed -i /log_line_prefix\ =/c\log_line_prefix\ =\ "'$log_line_prefix'"   $mDATA/postgresql.conf
sed -i 's/#host/host/'  $mDATA/pg_hba.conf

#restart master
pg_ctl restart
psql -c "alter user $USER password '123456'"
psql -c "select pg_start_backup('label');"
#copy data
cp -r $mDATA/* $sDATA
psql -c "select pg_stop_backup();"

#modify standby port
sed -i /port\ =/c\port\ =\ $sport                       $sDATA/postgresql.conf
sed -i /hot_standby\ =/c\hot_standby\ =\ $shot_standby   $sDATA/postgresql.conf
rm -rf $sDATA/postmaster.pid
#recovery.conf
cp $PGHOME/share/recovery.conf.sample  $sDATA/recovery.conf
sed -i /standby_mode/c\standby_mode=on $sDATA/recovery.conf
sed -i /primary_conninfo\ =/c\primary_conninfo="'host=127.0.0.1 port=5432 user=$USER password=123456'"   $sDATA/recovery.conf
sed -i /trigger_file/c\trigger_file="'$sDATA/trigger'" $sDATA/recovery.conf


pg_ctl start -D $sDATA

參考