postgres 数据库物理备份与还原
*备份$PGDATA、归档文件、表空间目录，一般适用于小版本的恢复，不能跨平台。
*数据库需要开启归档模式
本文主要介绍通过远程主机备份与还原postgres数据库
配置环境
主机名 IP地址 角色
qxy 192.168.40.170 备份数据库
qxy1 192.168.40.171 备份主机
1、数据库开启归档模式(qxy主机)
创建归档目录
mkdir -p /spark/pgsql/data_test/arch
[postgres@qxy ~]$
[postgres@qxy ~]$ mkdir -p /spark/pgsql/data_test/arch
[postgres@qxy ~]$
配置归档命令
vi $PGDATA/postgresql.conf
archive_mode = on
archive_command = 'DATE=`date +%Y%m%d`; DIR="/spark/pgsql/data_test/arch/$DATE"; (test -d $DIR || mkdir -p $DIR) && cp %p $DIR/%f'
[postgres@qxy data_test]$ cat postgresql.conf | grep archive_
archive_mode = on # enables archiving; off, on, or always
archive_command = ''DATE=`date +%Y%m%d`; DIR="/spark/pgsql/data_test/arch/$DATE"; (test -d $DIR || mkdir -p $DIR) && cp %p $DIR/%f'' # command to use to archive a logfile segment
%p 表示xlog文件名$PGDATA的相对路径, 如pg_xlog/000000010000000000000002
%f 表示xlog文件名, 000000010000000000000002
配置日志模
vi $PGDATA/postgresql.conf
wal_level = hot_standby或者archive模式
[postgres@qxy data_test]$ cat postgresql.conf | grep wal_level
wal_level = hot_standby # minimal, replica, or logical
[postgres@qxy data_test]$
重启数据库
[postgres@qxy data_test]$ pg_ctl restart -D /spark/pgsql/data_test/ -m fast
waiting for server to shut down..... done
server stopped
server starting
[postgres@qxy data_test]$ LOG: redirecting log output to logging collector process
HINT: Future log output will appear in directory "pg_clog".
[postgres@qxy data_test]$
[postgres@qxy data_test]$
[postgres@qxy data_test]$
测试归档是否正常
checkpoint;
select pg_switch_xlog();
[postgres@qxy arch]$ pwd
/spark/pgsql/data_test/arch
[postgres@qxy arch]$ ls -ltr
total 4
drwx------. 2 postgres postgres 4096 Jun 9 22:31 20180609
[postgres@qxy arch]$ cd 20180609/
[postgres@qxy 20180609]$ ls -ltr
total 32768
-rw-------. 1 postgres postgres 16777216 Jun 9 22:31 000000010000000000000002
-rw-------. 1 postgres postgres 16777216 Jun 9 22:31 000000010000000000000003
[postgres@qxy 20180609]$ 2、创建一个测试表空间(qxy主机)
创建表空间目录
mkdir -p /spark/pgsql/tablespace/tbs01
create tablespace tblsp1 location '/spark/pgsql/tablespace/tbs01';
create tablespace tblsp2 location '/spark/pgsql/tablespace/tbs01';
[postgres@qxy ~]$
[postgres@qxy ~]$ mkdir -p /spark/pgsql/tablespace/tbs01
[postgres@qxy ~]$
[postgres@qxy ~]$ psql -p 5433
psql.bin (9.6.4)
Type "help" for help.
postgres=# create tablespace tblsp1 location '/spark/pgsql/tablespace/tbs01';
CREATE TABLESPACE
postgres=# create tablespace tblsp2 location '/spark/pgsql/tablespace/tbs01';
ERROR: directory "/spark/pgsql/tablespace/tbs01/PG_9.6_201608131" already in use as a tablespace <=====提示已经存在一个目录，如果还需要新建一个表空间，需要在创建一个目录
postgres=#
postgres=#
查看创建的表空间
postgres=# \db
List of tablespaces
Name | Owner | Location
------------+----------+-------------------------------
pg_default | postgres |
pg_global | postgres |
tblsp1 | postgres | /spark/pgsql/tablespace/tbs01
(3 rows)
postgres=# select * from pg_tablespace;
spcname | spcowner | spcacl | spcoptions
------------+----------+--------+------------
pg_default | 10 | |
pg_global | 10 | |
tblsp1 | 10 | |
(3 rows)
postgres=#
表空间使用
postgres=#
postgres=# create table test (id int) tablespace tblsp1;
CREATE TABLE
postgres=#
postgres=#
设置默认表空间
如果我们不建立自己的表空间，建立表的时候，也不指定表空间，那么表默认会放入缺省表空间里，即pg_default
postgres=# \l+
List of databases
Name | Owner | Encoding | Collate | Ctype | Access privileges | Size | Tablespace | Description
-----------+----------+----------+---------+-------+-----------------------+---------+------------+--------------------------------------------
postgres | postgres | UTF8 | C | C | | 7071 kB | pg_default | default administrative connection database
qxy | postgres | UTF8 | C | C | | 6993 kB | pg_default |
template0 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | unmodifiable empty database
| | | | | postgres=CTc/postgres | | |
template1 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | default template for new databases
| | | | | postgres=CTc/postgres | | |
(4 rows)
postgres=# alter database qxy set tablespace tblsp1;
ALTER DATABASE
postgres=# \c qxy scott
You are now connected to database "qxy" as user "scott".
qxy=# create table test (id int) tablespace tblsp1;
CREATE TABLE
qxy=#
qxy=#
表空间对应与文件系统
[postgres@qxy pg_tblspc]$ pwd
/spark/pgsql/data_test/pg_tblspc
[postgres@qxy pg_tblspc]$ ls -ltr
total 0
lrwxrwxrwx. 1 postgres postgres 29 Jun 9 22:36 32773 -> /spark/pgsql/tablespace/tbs01
[postgres@qxy pg_tblspc]$
初始化cluster时默认的表空间说明
database cluster初始化时创建两个默认表空间
pg_global –存放系统字典表
pg_default –template0,template1默认表空间,所以create database不明确指定表空间都使用此表空间3、使用过pg_basebackup, 流复制协议备份（qxy主机）
创建replication权限的角色, 或者超级用户的角色
create role replica nosuperuser replication login connection limit 32 encrypted password 'replica';
postgres=# create role replica nosuperuser replication login connection limit 32 encrypted password 'replica';
CREATE ROLE
postgres=#
配置pg_hba.conf
添加下面一行
host replication replica 192.168.40.1/24 md5 <====其中192.168.40.1/24 代表192.168.40字段的IP都可以访问，replication代表具有流复制权限
[postgres@qxy data_test]$ cat pg_hba.conf | grep -v '#' | grep -ni 'replication'
8:host replication replica 192.168.40.1/24 md5
[postgres@qxy data_test]$
[postgres@qxy data_test]$ pg_ctl reload -D /spark/pgsql/data_test/
server signaled
[postgres@qxy data_test]$
[postgres@qxy data_test]$
配置 postgresql.conf
把listen_addresses = 'localhost'修改为listen_addresses = '*' <=====默认只监听本机，修改为* 表示为监听所有IP的发起的连接
[postgres@qxy data_test]$ cat postgresql.conf | grep listen
listen_addresses = '*' # what IP address(es) to listen on;
[postgres@qxy data_test]$
重启数据库生效
[postgres@qxy data_test]$ pg_ctl restart -D /spark/pgsql/data_test/
waiting for server to shut down.... done
server stopped
server starting
[postgres@qxy data_test]$ LOG: redirecting log output to logging collector process
HINT: Future log output will appear in directory "pg_clog".
[postgres@qxy data_test]$
远端机器连接测试（qxy1主机）
[postgres@QXY1 ~]$ hostname
QXY1.localdomain
[postgres@QXY1 ~]$
[postgres@QXY1 ~]$ psql -h 192.168.40.170 -p 5433 -d postgres -U replica --password
Password for user replica:
psql.bin (9.6.4)
Type "help" for help.
postgres=>
postgres=> \l+
List of databases
Name | Owner | Encoding | Collate | Ctype | Access privileges | Size | Tablespace | Description
-----------+----------+----------+---------+-------+-----------------------+---------+------------+--------------------------------------------
postgres | postgres | UTF8 | C | C | | 7071 kB | pg_default | default administrative connection database
qxy | postgres | UTF8 | C | C | | 7103 kB | tblsp1 |
template0 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | unmodifiable empty database
| | | | | postgres=CTc/postgres | | |
template1 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | default template for new databases
| | | | | postgres=CTc/postgres | | |
(4 rows)
postgres=> 4、备份存放主机创建备份目录(QXY1主机)
[postgres@QXY1 pgbak]$
[postgres@QXY1 pgbak]$ mkdir -p /home/postgres/pgbak
[postgres@QXY1 pgbak]$ cd /home/postgres/pgbak
pg_basebackup -F t -x -D ./`date +%F` -h 192.168.40.170 -p 5433 -U replica
[postgres@QXY1 pgbak]$ ls -ltr
total 0
[postgres@QXY1 pgbak]$ pwd
/home/postgres/pgbak
[postgres@QXY1 pgbak]$ pg_basebackup -F t -x -D ./`date +%F` -h 192.168.40.170 -p 5433 -U replica
Password:
pg_basebackup: could not connect to server: FATAL: number of requested standby connections exceeds max_wal_senders (currently 0) <====源库重新修改max_wal_senders大于0即可
[postgres@QXY1 pgbak]$ pg_basebackup -F t -x -D ./`date +%F` -h 192.168.40.170 -p 5433 -U replica
Password:
[postgres@QXY1 pgbak]$ ls
2018-06-10
[postgres@QXY1 pgbak]$ du -sh 2018-06-10/
157M 2018-06-10/
[postgres@QXY1 pgbak]$
[postgres@QXY1 2018-06-10]$ ls -ltr
total 160328
-rw-r--r--. 1 postgres postgres 7416320 Jun 10 07:16 32773.tar <=====代表是表空间
-rw-r--r--. 1 postgres postgres 156755456 Jun 10 07:17 base.tar <=====就是$PGDATA
postgres@QXY1 2018-06-10]$ tar -tf 32773.tar <====查看32773.tar里面的内容，其实就是源库的表空间内容
PG_9.6_201608131/
PG_9.6_201608131/12404/
PG_9.6_201608131/12404/32775
PG_9.6_201608131/24576/
PG_9.6_201608131/24576/12244_fsm
PG_9.6_201608131/24576/2609_fsm
PG_9.6_201608131/24576/2839
PG_9.6_201608131/24576/3256_vm
PG_9.6_201608131/24576/2684
PG_9.6_201608131/24576/3468
PG_9.6_201608131/24576/12249_vm
PG_9.6_201608131/24576/549
PG_9.6_201608131/24576/12239
PG_9.6_201608131/24576/3602_fsm
PG_9.6_201608131/24576/3455
PG_9.6_201608131/24576/2835
PG_9.6_201608131/24576/2618
PG_9.6_201608131/24576/2666
PG_9.6_201608131/24576/1255
PG_9.6_201608131/24576/2663
PG_9.6_201608131/24576/3576_vm
PG_9.6_201608131/24576/2650
PG_9.6_201608131/24576/2606_vm
PG_9.6_201608131/24576/2618_fsm
PG_9.6_201608131/24576/3574
PG_9.6_201608131/24576/2667
。
。
。 <=====省略
。
[postgres@QXY1 2018-06-10]$ tar -tf base.tar | more
backup_label <======多了一个backup_level，记录了备份时间点的相关信息
tablespace_map
pg_dynshmem/
pg_ident.conf
pg_logical/
pg_logical/snapshots/
pg_logical/mappings/
pg_stat_tmp/
pg_clog/
pg_clog/postgresql-2018-06-09_230730.log
pg_clog/postgresql-2018-06-09_223028.log
pg_clog/postgresql-2018-06-09_231641.log
pg_clog/postgresql-2018-06-09_223107.log
pg_clog/postgresql-2018-06-09_133147.log
pg_clog/postgresql-2018-06-09_230203.log
pg_clog/0000
pg_clog/postgresql-2018-06-09_231200.log
pg_multixact/
pg_multixact/members/
pg_multixact/members/0000
pg_multixact/offsets/
pg_multixact/offsets/0000
postgresql.conf
pg_commit_ts/
源库表空间内容信息，可以看到和 32773.tar里面的内容一样
[postgres@qxy PG_9.6_201608131]$ cd
[postgres@qxy ~]$ cd /spark/pgsql/data_test/
[postgres@qxy data_test]$ cd pg_tblspc/
[postgres@qxy pg_tblspc]$ ls -ltr
total 0
lrwxrwxrwx. 1 postgres postgres 29 Jun 9 22:36 32773 -> /spark/pgsql/tablespace/tbs01
[postgres@qxy pg_tblspc]$ cd /spark/pgsql/tablespace/tbs01
[postgres@qxy tbs01]$ ls -ltr
total 4
drwx------. 4 postgres postgres 4096 Jun 9 22:44 PG_9.6_201608131
[postgres@qxy tbs01]$ cd PG_9.6_201608131/
[postgres@qxy PG_9.6_201608131]$ ls -ltr
total 8
drwx------. 2 postgres postgres 4096 Jun 9 22:40 12404
drwx------. 2 postgres postgres 4096 Jun 9 23:16 24576
[postgres@qxy PG_9.6_201608131]$
注：
因为测试时归档目录存放在了/spark/pgsql/data_test/下面，所以备份的时候已经顺带着把归档备份走了，如果备份目录是另外独立的，需要手动copy。5、源库模拟删除(qxy主机)
[postgres@qxy ~]$ psql -p 5433
psql.bin (9.6.4)
Type "help" for help.
postgres=# \l+
List of databases
Name | Owner | Encoding | Collate | Ctype | Access privileges | Size | Tablespace | Description
-----------+----------+----------+---------+-------+-----------------------+---------+------------+--------------------------------------------
postgres | postgres | UTF8 | C | C | | 7071 kB | pg_default | default administrative connection database
qxy | postgres | UTF8 | C | C | | 7103 kB | tblsp1 |
template0 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | unmodifiable empty database
| | | | | postgres=CTc/postgres | | |
template1 | postgres | UTF8 | C | C | =c/postgres +| 6953 kB | pg_default | default template for new databases
| | | | | postgres=CTc/postgres | | |
(4 rows)
postgres=# \d t
Table "public.t"
Column | Type | Modifiers
--------+---------+-----------
id | integer |
postgres=# select * from t;
id
----
10
10
10
10
10
10
(6 rows)
postgres=# insert into t values (10); <====重新插入一条记录，这条记录备份的时候是不存在的，等下要测试通过备份集加上归档是否可以还原改行数据
INSERT 0 1
postgres=#
postgres=# checkpoint;
CHECKPOINT
postgres=# select pg_switch_xlog();
pg_switch_xlog
----------------
0/A0005D8
(1 row)
postgres=# \q
[postgres@qxy ~]$
--删除数据库
[postgres@qxy ~]$ rm -rf /spark/pgsql/data_test/*
[postgres@qxy ~]$ rm -rf /spark/pgsql/tablespace/tbs01/*
[postgres@qxy ~]$
[postgres@qxy ~]$
[postgres@qxy ~]$ 6、通过备份集还原源端数据库
（从备份库(QXY1)把备份结果集scp到源库）
[postgres@QXY1 2018-06-10]$ scp 32773.tar [email protected]:/spark/pgsql/tablespace/tbs01/
[email protected]'s password:
32773.tar 100% 7243KB 7.1MB/s 00:00
[postgres@QXY1 2018-06-10]$ scp base.tar [email protected]:/spark/pgsql/data_test/
[email protected]'s password:
base.tar 100% 149MB 149.5MB/s 00:01
[postgres@QXY1 2018-06-10]$
源库解压(qxy主机执行)
[postgres@qxy data_test]$ tar -xvf base.tar
[postgres@qxy data_test]$ ls -ltr
total 153208
-rw-------. 1 postgres postgres 88 Jun 7 15:30 postgresql.auto.conf
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_twophase
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_subtrans
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_snapshots
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_serial
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_replslot
drwx------. 4 postgres postgres 4096 Jun 7 15:30 pg_multixact
drwx------. 4 postgres postgres 4096 Jun 7 15:30 pg_logical
-rw-------. 1 postgres postgres 1636 Jun 7 15:30 pg_ident.conf
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_dynshmem
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_commit_ts
-rw-------. 1 postgres postgres 4 Jun 7 15:30 PG_VERSION
drwxrwxr-x. 3 postgres postgres 4096 Jun 9 22:31 arch
drwx------. 2 postgres postgres 4096 Jun 9 22:36 pg_tblspc
drwx------. 5 postgres postgres 4096 Jun 9 22:44 base
-rw-------. 1 postgres postgres 4604 Jun 9 23:09 pg_hba.conf
-rw-------. 1 postgres postgres 22462 Jun 9 23:16 postgresql.conf
drwx------. 2 postgres postgres 4096 Jun 9 23:16 pg_stat
drwx------. 2 postgres postgres 4096 Jun 9 23:16 pg_notify
drwx------. 2 postgres postgres 4096 Jun 9 23:16 pg_clog
drwx------. 2 postgres postgres 4096 Jun 9 23:16 global
-rw-------. 1 postgres postgres 36 Jun 9 23:16 tablespace_map
drwx------. 2 postgres postgres 4096 Jun 9 23:16 pg_stat_tmp
-rw-------. 1 postgres postgres 206 Jun 9 23:16 backup_label <====记录了备份时检查点相关的信息
-rw-r--r--. 1 postgres postgres 156755456 Jun 9 23:35 base.tar
drwx------. 3 postgres postgres 4096 Jun 9 23:36 pg_xlog
[postgres@qxy data_test]$ cat backup_label
START WAL LOCATION: 0/9000028 (file 000000010000000000000009)
CHECKPOINT LOCATION: 0/9000060
BACKUP METHOD: streamed
BACKUP FROM: master
START TIME: 2018-06-09 23:16:50 GMT
LABEL: pg_basebackup base backup
[postgres@qxy data_test]$
[postgres@qxy tbs01]$ tar -xvf 32773.tar <=====解压表空间
[postgres@qxy tbs01]$ cd PG_9.6_201608131/
[postgres@qxy PG_9.6_201608131]$ ls
12404 24576
[postgres@qxy PG_9.6_201608131]$ ls -ltr
total 8
drwx------. 2 postgres postgres 4096 Jun 9 22:40 12404
drwx------. 2 postgres postgres 4096 Jun 9 23:16 24576
[postgres@qxy PG_9.6_201608131]$ pwd
/spark/pgsql/tablespace/tbs01/PG_9.6_201608131
[postgres@qxy PG_9.6_201608131]$
7、源端数据库配置recovery.conf（qxy主机执行）
第5步的删除postgres数据库之前，插入了一条记录，所以删除之前时7条数据记录，但是备份postgres数据库时，测试表T里面只有6条记录，现在就是需要通过备份集加上归档来恢复最后一条记录。所以这里需要配置recovery.conf
copy $PGHOME/share/postgres/recovery.conf.sample 到 /spark/pgsql/data_test
mv recovery.conf.sample recovery.conf
配置下面这行
restore_command = 'cp /spark/pgsql/data_test/arch/20180610/%f %p' 8、还原数据库(qxy主机执行)
重新启动数据库
[postgres@qxy ~]$ pg_ctl start -D /spark/pgsql/data_test/
server starting
[postgres@qxy ~]$ LOG: redirecting log output to logging collector process
HINT: Future log output will appear in directory "pg_clog".
[postgres@qxy ~]$
[postgres@qxy ~]$ ps -ef | grep postgres
root 3413 3269 0 22:24 pts/2 00:00:00 su - postgres
postgres 3414 3413 0 22:24 pts/2 00:00:00 -bash
root 3679 3294 0 22:36 pts/3 00:00:00 su - postgres
postgres 3680 3679 0 22:36 pts/3 00:00:00 -bash
root 6372 3245 0 23:36 pts/1 00:00:00 su - postgres
postgres 6373 6372 0 23:36 pts/1 00:00:00 -bash
postgres 6714 1 0 23:42 pts/1 00:00:00 /spark/pgsql/pgsql/bin/postgres -D /spark/pgsql/data_test
postgres 6715 6714 0 23:42 ? 00:00:00 postgres: logger process
postgres 6717 6714 0 23:42 ? 00:00:00 postgres: checkpointer process
postgres 6718 6714 0 23:42 ? 00:00:00 postgres: writer process
postgres 6719 6714 0 23:42 ? 00:00:00 postgres: wal writer process
postgres 6720 6714 0 23:42 ? 00:00:00 postgres: autovacuum launcher process
postgres 6721 6714 0 23:42 ? 00:00:00 postgres: archiver process
postgres 6722 6714 0 23:42 ? 00:00:00 postgres: stats collector process
postgres 6724 6373 10 23:42 pts/1 00:00:00 ps -ef
postgres 6725 6373 0 23:42 pts/1 00:00:00 grep postgres
[postgres@qxy ~]$ psql -p 5432
psql.bin: could not connect to server: No such file or directory
Is the server running locally and accepting
connections on Unix domain socket "/tmp/.s.PGSQL.5432"?
[postgres@qxy ~]$ psql -p 5433
psql.bin (9.6.4)
Type "help" for help.
postgres=# select * from t;
id
----
10
10
10
10
10
10
10 <========可以看到，最后插入的一条记录也还原回来
(7 rows)
恢复时部分日志如下：
LOG: database system was shut down at 2018-06-10 00:28:16 GMT
LOG: starting archive recovery
LOG: restored log file "000000030000000000000010" from archive
LOG: restored log file "00000003.history" from archive
LOG: restored log file "00000004.history" from archive
LOG: redo starts at 0/10000028
LOG: consistent recovery state reached at 0/100000F8
LOG: restored log file "000000040000000000000011" from archive
LOG: restored log file "000000050000000000000012" from archive
LOG: restored log file "000000050000000000000013" from archive
LOG: invalid record length at 0/14000098: wanted 24, got 0
LOG: redo done at 0/14000028
LOG: last completed transaction was at log time 2018-06-10 00:26:31.331415+00
LOG: restored log file "00000005.history" from archive
LOG: archive recovery complete
[postgres@qxy data_test]$ ls -ltr
total 153224
-rw-------. 1 postgres postgres 88 Jun 7 15:30 postgresql.auto.conf
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_twophase
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_subtrans
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_snapshots
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_serial
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_replslot
drwx------. 4 postgres postgres 4096 Jun 7 15:30 pg_multixact
drwx------. 4 postgres postgres 4096 Jun 7 15:30 pg_logical
-rw-------. 1 postgres postgres 1636 Jun 7 15:30 pg_ident.conf
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_dynshmem
drwx------. 2 postgres postgres 4096 Jun 7 15:30 pg_commit_ts
-rw-------. 1 postgres postgres 4 Jun 7 15:30 PG_VERSION
drwxrwxr-x. 3 postgres postgres 4096 Jun 9 22:31 arch
drwx------. 5 postgres postgres 4096 Jun 9 22:44 base
-rw-------. 1 postgres postgres 4604 Jun 9 23:09 pg_hba.conf
-rw-------. 1 postgres postgres 22462 Jun 9 23:16 postgresql.conf
-rw-------. 1 postgres postgres 36 Jun 9 23:16 tablespace_map.old
-rw-------. 1 postgres postgres 206 Jun 9 23:16 backup_label.old <========== 恢复完成后，该文件变成了old
-rw-r--r--. 1 postgres postgres 156755456 Jun 9 23:35 base.tar
drwx------. 2 postgres postgres 4096 Jun 9 23:42 pg_tblspc
-rw-r--r--. 1 postgres postgres 5720 Jun 9 23:45 recovery.done <========== recovery.conf变成了done
-rw-------. 1 postgres postgres 71 Jun 9 23:45 postmaster.pid
drwx------. 2 postgres postgres 4096 Jun 9 23:45 pg_notify
-rw-------. 1 postgres postgres 62 Jun 9 23:45 postmaster.opts
drwx------. 2 postgres postgres 4096 Jun 9 23:45 pg_clog
drwx------. 2 postgres postgres 4096 Jun 9 23:45 pg_stat
drwx------. 3 postgres postgres 4096 Jun 9 23:45 pg_xlog
drwx------. 2 postgres postgres 4096 Jun 9 23:45 global
drwx------. 2 postgres postgres 4096 Jun 9 23:50 pg_stat_tmp
[postgres@qxy data_test]$ 9、其他方式备份
[postgres@qxy pg_clog]$ psql -p 5433
psql.bin (9.6.4)
Type "help" for help.
postgres=# \df *backup*
List of functions
Schema | Name | Result data type | Argument data types | Type
------------+----------------------+--------------------------+----------------------------------------------------------------------------+--------
pg_catalog | pg_backup_start_time | timestamp with time zone | | normal
pg_catalog | pg_is_in_backup | boolean | | normal
pg_catalog | pg_start_backup | pg_lsn | label text, fast boolean DEFAULT false, exclusive boolean DEFAULT true | normal
pg_catalog | pg_stop_backup | pg_lsn | | normal
pg_catalog | pg_stop_backup | SETOF record | exclusive boolean, OUT lsn pg_lsn, OUT labelfile text, OUT spcmapfile text | normal
(5 rows)
postgres=#
首先执行pg_start_backup函数，然后就可以手动copy备份，备份的目录包括$PGDATA、归档、表空间。
如：
postgres=# select pg_start_backup(now()::text);
pg_start_backup
-----------------
0/15000060
(1 row)
postgres=# select pg_is_in_backup();
pg_is_in_backup
-----------------
t
(1 row)
postgres=#
copy结束之后，执行pg_stop_backup
postgres=# select pg_stop_backup();
NOTICE: pg_stop_backup complete, all required WAL segments have been archived
pg_stop_backup
----------------
0/15000168
(1 row)
postgres=# select pg_is_in_backup();
pg_is_in_backup
-----------------
f
(1 row)
postgres=#
注意：
copy过程中的归档也需要备份。