1 介绍
PostgresSQL在9.0版本实现的复制功能只能进行异步复制，在PostgreSQL 9.1版本中可以支持同步复制。
PostgreSQL流复制是基于WAL日志传输实现的：主库发送WAL日志，备库接收WAL日志并进行回放。
流复制的3个主要进程:
- 主库的walsender进程：负责发送WAL日志给备库。
- 备库的walreceiver进程：负责主库发送的WAL日志。
- 备库的startup进程：重放接收的WAL日志。
主从之间是怎么通信的？
（1）primary端 后端进程通过执行函数XLogInsert()和XLogFlush()，将WAL数据写入并刷新到WAL段文件中。
（2）primary端 WAL发送器进程将写入WAL段文件的WAL数据发送到WAL接收器进程。
（3）primary端 在发送WAL数据之后，后端进程继续等待来自备库的ACK响应。更确切地说，后端进程通过执行内部函数SyncRepWaitForLSN()来获取锁存器，并等待它被释放。
（4）standby端 上的WAL接收器通过write()系统函数调用，将接收到的WAL数据写入备库的WAL段，并向WAL发送器返回ACK响应。
（5）standby端 WAL接收器通过调用（如fsync()函数）将WAL数据刷新到WAL段中，向WAL发送器返回另一个ACK响应，并通知启动进程相关WAL数据已更新。
（6）standby端 启动进程重放已写入WAL段的WAL数据。
（7）primary端 WAL发送器在收到来自WAL接收器的ACK响应后释放后端进程的锁存器，然后后端进程完成commit或abort动作。锁存器释放的时间取决于参数synchronous_commit。如果它是‘on’（默认），那么当接收到步骤（5）的ACK时，锁存器被释放；如果它是'remote_write'，当接收到步骤（4）的ACK时，其被释放。
每个ACK响应将备库的内部信息通知给主库，包含以下4个项目：
- 已写入最新WAL数据的LSN位置。
- 已刷新最新WAL数据的LSN位置。
- 启动进程已经重放最新的WAL数据的LSN。
- 发送此响应的时间戳。
2 安装PostgreSQL 14.5
CentOS 7.6
PostgreSQL 14.5
安装步骤请参考：PostgreSQL 14.5 for CentOS安装
3 配置复制流（同步复制）
同步流复制primary数据库要求wal日志写入standby数据库commit后才会返回成功，所以当standby与primary断开时，primary端会hang住。为了解决这个问题需要与primary配置至少两个standby，确保primary数据库的wal日志写入任意一个standby数据库并commit。
3.1 配置“复制槽”
standby端如果长时间停机，重启后standby可能因缺少相应的WAL日志无法连接primary。此时可以通过启用max_replication_slots参数启用复制槽来解决此问题。
primary端实例会一直保留预写日志（WAL）文件，直到所有备库所需的插槽都确认已接收到特定段为止。只有完成此操作后，主库实例才会移除相应的WAL文件。
Primary端创建复制槽
SELECT * FROM pg_create_physical_replication_slot('pg_primary');
查看复制槽
postgres=# select * from pg_replication_slots;
slot_name | plugin | slot_type | datoid | database | active | active_pid | xmin | catalog_xmin | restart_lsn | confirmed_flush_lsn
------------+--------+-----------+--------+----------+--------+------------+------+--------------+-------------+---------------------
pg_primary | | physical | | | t | 116077 | | | 89/A40003E0 | 3.2 开启归档
开启归档模式
创建归档目录
[postgres]# mkdir $PGDATA/arch_log配置归档脚本
[postgres]# vi $PGDATA/archive.sh
cp --preserve=timestamps $1 $PGDATA/arch_log/$2 ; find $PGDATA/arch_log -type f -mtime +30 | xargs rm -fr;修改归档相关参数
[postgres]# vi $PGDATA/postgresql.conf
archive_mode = on
archive_command = '/bin/bash archive.sh %p %f'
archive_timeout = 1800重启postgres使归档参数生效
[postgres]# pg_ctl restart查看参数是否已经生效
postgres=# select name,setting from pg_settings where name in ('archive_mode','archive_command','archive_timeout');
name | setting
-----------------+----------------------------
archive_command | /bin/bash archive.sh %p %f
archive_mode | on
archive_timeout | 1800手动归档
postgres=# select pg_switch_xlog();
pg_switch_wal
---------------
0/4000160查看归档目录是否已存在归档文件
[postgres]# ls $PGDATA/arch_log
0000000100000000000000043.3 primary端配置
创建复制用户
postgres=# create role repl login replication encrypted password 'Your_passwd';
CREATE ROLE配置权限配置文件
[postgres]# vi pg_hba.conf
添加如下：
host replication repl 0.0.0.0/0 md5配置参数（这个必须配置）
[postgres]# vi postgresql.conf
listen_addresses = '*'
wal_level = replica
hot_standby = on #指定备库可读模式
synchronous_commit = on #配置同步模式
synchronous_standby_names = 'host72,host73' #只要primary端wal传到任意一台standby并commit就OK（此处是application_name）重启PG使参数生效
[postgres]# pg_ctl restart3.4 配置standby端
删除standby端数据
[postgres]# pg_ctl stop
[postgres]# rm -fr $PGDATA/*复制primary端到standby端，在$PGDATA路径下会生成standby.signal文件
[postgres]# pg_basebackup -h 192.168.2.221 -U repl -D $PGDATA -X stream -P -R
配置主从参数
[postgres]# vi $PGDATA/postgresql.auto.conf
primary_conninfo = 'application_name=host72 user=repl password=Your_passwd channel_binding=disable host=192.168.1.71 port=5432 sslmode=disable sslcompression=0 sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=disable krbsrvname=postgres target_session_attrs=any'
[postgres]# vi $PGDATA/postgresql.conf
#primary_conninfo = 'host=192.168.1.71 port=5432 user=repl password=Your_passwd' #我前面配置了postgresql.auto.conf此处不配置
recovery_target_timeline = 'latest'
primary_slot_name = 'pg_primary'
hot_standby = on #配置standby端可读3.5 启动流复制
standby端启动postgres
[postgres]# pg_ctl start
primary端查看流复制详情
postgres=# select * from pg_stat_replication;
-[ RECORD 1 ]----+------------------------------
pid | 22930
usesysid | 16384
usename | repl
application_name | host72
client_addr | 192.168.1.72
client_hostname |
client_port | 54718
backend_start | 2022-09-22 21:42:14.120187+08
backend_xmin |
state | streaming
sent_lsn | 0/6F0003B8
write_lsn | 0/6F0003B8
flush_lsn | 0/6F0003B8
replay_lsn | 0/6F0003B8
write_lag | 00:00:00.000336
flush_lag | 00:00:00.000336
replay_lag | 00:00:00.000336
sync_priority | 1
sync_state | sync
reply_time | 2022-09-22 21:42:14.229005+08
-[ RECORD 1 ]----+------------------------------
pid | 22930
usesysid | 16384
usename | repl
application_name | host72
client_addr | 192.168.1.73
client_hostname |
client_port | 54718
backend_start | 2022-09-22 21:42:14.120187+08
backend_xmin |
state | streaming
sent_lsn | 0/6F0003B8
write_lsn | 0/6F0003B8
flush_lsn | 0/6F0003B8
replay_lsn | 0/6F0003B8
write_lag | 00:00:00.000336
flush_lag | 00:00:00.000336
replay_lag | 00:00:00.000336
sync_priority | 1
sync_state | sync
reply_time | 2022-09-22 21:42:14.229005+084 监控复制流
Primary端查看已连接的Standby端信息，如果为空说明复制流异常
postgres=# \x on;
Expanded display is on.
postgres=# select * from pg_stat_replication;
-[ RECORD 1 ]----+------------------------------
pid | 4007
usesysid | 16384
usename | repl
application_name | walreceiver
client_addr | 192.168.1.72
client_hostname | host72
client_port | 34596
backend_start | 2022-09-21 17:47:09.629914+08
backend_xmin |
state | streaming
sent_location | 1/F4000060
write_location | 1/F4000060
flush_location | 1/F4000060
replay_location | 1/F4000060
sync_priority | 0
sync_state | sync
----+------------------------------
pid | 116097
usesysid | 16385
usename | repl
application_name | walreceiver
client_addr | 192.168.1.73
client_hostname |
client_port | 45984
backend_start | 2021-12-20 16:17:58.493769+08
backend_xmin |
state | streaming
sent_location | 89/A2BE8E10
write_location | 89/A2BE8E10
flush_location | 89/A2BE8E10
replay_location | 89/A2BE8E10
sync_priority | 0
sync_state | sync
Standby端查看是否处于recovery状态，值为“t”说明是当前数据库是备库，值为“f”说明当前数据库是主库（也可能复制流出现异常出现双主）
postgres=# select pg_is_in_recovery();
pg_is_in_recovery
-------------------
t
通过pg_controldata命令查看Database cluster state，值为“in archive recovery”说明是当前数据库是备库，值为“in production”说明当前数据库是主库
[postgres@host72 postgres]$ pg_controldata $DATA
pg_control version number: 960
Catalog version number: 201608131
Database system identifier: 7122063347532860770
Database cluster state: in archive recovery
pg_control last modified: Wed 21 Sep 2022 05:52:09 PM CST
Latest checkpoint location: 1/F4000098
Prior checkpoint location: 1/F0000060
Latest checkpoint's REDO location: 1/F4000060
Latest checkpoint's REDO WAL file: 00000001000000010000003D
Latest checkpoint's TimeLineID: 1
Latest checkpoint's PrevTimeLineID: 1
Latest checkpoint's full_page_writes: on
Latest checkpoint's NextXID: 0:1762
Latest checkpoint's NextOID: 24576
Latest checkpoint's NextMultiXactId: 1
Latest checkpoint's NextMultiOffset: 0
Latest checkpoint's oldestXID: 1753
Latest checkpoint's oldestXID's DB: 1
Latest checkpoint's oldestActiveXID: 1762
Latest checkpoint's oldestMultiXid: 1
Latest checkpoint's oldestMulti's DB: 1
Latest checkpoint's oldestCommitTsXid:0
Latest checkpoint's newestCommitTsXid:0
Time of latest checkpoint: Wed 21 Sep 2022 05:51:43 PM CST
Fake LSN counter for unlogged rels: 0/1
Minimum recovery ending location: 1/F4000108
Min recovery ending loc's timeline: 1
Backup start location: 0/0
Backup end location: 0/0
End-of-backup record required: no
wal_level setting: replica
wal_log_hints setting: on
max_connections setting: 500
max_worker_processes setting: 8
max_prepared_xacts setting: 0
max_locks_per_xact setting: 64
track_commit_timestamp setting: off
Maximum data alignment: 8
Database block size: 8192
Blocks per segment of large relation: 131072
WAL block size: 8192
Bytes per WAL segment: 67108864
Maximum length of identifiers: 64
Maximum columns in an index: 32
Maximum size of a TOAST chunk: 1996
Size of a large-object chunk: 2048
Date/time type storage: 64-bit integers
Float4 argument passing: by value
Float8 argument passing: by value
Data page checksum version: 0
在standby查看wal接收情况
postgres=# \x on;
Expanded display is on.
postgres=# select * from pg_stat_wal_receiver;
-[ RECORD 1 ]---------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
pid | 34474
status | streaming
receive_start_lsn | 0/70000000
receive_start_tli | 2
written_lsn | 0/70000250
flushed_lsn | 0/70000250
received_tli | 2
last_msg_send_time | 2022-09-22 21:55:55.905271+08
last_msg_receipt_time | 2022-09-22 21:55:55.904179+08
latest_end_lsn | 0/70000250
latest_end_time | 2022-09-22 21:55:25.866616+08
slot_name | pg_primary
sender_host | 192.168.1.71
sender_port | 5432
conninfo | user=repl password=******** channel_binding=disable dbname=replication host=192.168.1.71 port=5432 application_name=host72 fallback_application_name=walreceiver sslmode=disable sslcompression=0 sslsni=1 ssl_min_protocol_version=TLSv1.2 gssencmode=disable krbsrvname=postgres target_session_attrs=any
primary端查看备库落后主库wal的字节数，可以通过sysbench压力测试并观察下
postgres=# select pg_wal_lsn_diff(pg_current_wal_lsn(),replay_lsn) from pg_stat_replication;
pg_wal_lsn_diff
-----------------
0
