repmgr是一个对postgresql流复制进行管理以及自动故障转移的开源软件，大大的简化了PG流复制架构的管理。但是repmgr不具备提供VIP功能，只能借助keepalived实现VIP，并确保VIP运行在primary节点上。其不具备连接池功能，所以只是一个轻量级开源软件。下面详细介绍该架构的详细搭建过程。
先安装repmgr依赖的软件包，官方推荐使用yum安装如下组件和rpm包
yumcheck-update
yumgroupinstall "Development Tools"
yuminstall yum-utils openjade docbook-dtds docbook-style-dsssldocbook-style-xsl
yuminstall flex libselinux-devel libxml2-devel libxslt-developenssl-devel pam-devel readline-devel
下载repmgr并安装，当前最新的版本为5.2.1.
./configure&& make install
注意，确保pg_config在安装用户的环境变量PATH中，repmgr会默认安装到postgres的软件目录下。
首先完成流复制相关的参数修改和用户创建，并修改pg_hba.conf文件
createuser repluser with usperuser password '****';
createdatabase repmgr with owner='repluser';
--修改主备节点的pg_hba.conf文件
local replication repluser trust host replication repluser 127.0.0.1/32 trust host replication repluser 10.26.60.0/24 scram-sha-256 local repmgr repluser trust host repmgr repluser 127.0.0.1/32 trust host repmgr repluser 10.26.60.0/24 scram-sha-256 |
--修改.pgpass
spcl-pg250:5432:repmgr:repluser:Repl#2021
spcl-pg249:5432:repmgr:repluser:Repl#2021
--修改/etc/repmgr.conf
node_id=1
node_name='host01'
conninfo='host=host01user=repluser dbname=repmgr connect_timeout=2'
data_directory='/pgdata’
--注册主节点
repmgr-f etc/repmgr.conf primary register
--验证集群状态
repmgr-f etc/repmgr.conf cluster show
--创建从库的/etc/repmgr.conf文件
node_id=2
node_name='host02'
conninfo='host=host02user=repluser dbname=repmgr connect_timeout=2'
data_directory='/pgdata'
--进行从库搭建前的检测
[postgres@host02pgdata]$ repmgr -h host01 -U repluser -d repmgr -f/pgdata/repmgr.conf standby clone --dry-run
NOTICE:destination directory "/pgdata" provided
INFO:connecting to source node
DETAIL:connection string is: host=host01 user=repluser dbname=repmgr
DETAIL:current installation size is 31 MB
INFO:"repmgr" extension is installed in database "repmgr"
INFO:replication slot usage not requested; no replication slot will beset up for this standby
INFO:parameter "max_wal_senders" set to 32
NOTICE:checking for available walsenders on the source node (2 required)
INFO:sufficient walsenders available on the source node
DETAIL:2 required, 31 available
NOTICE:checking replication connections can be made to the source server (2required)
INFO:required number of replication connections could be made to thesource server
DETAIL:2 replication connections required
WARNING:data checksums are not enabled and "wal_log_hints" is "off"
DETAIL:pg_rewind requires "wal_log_hints" to be enabled
NOTICE:standby will attach to upstream node 1
HINT:consider using the -c/--fast-checkpoint option
INFO:all prerequisites for "standby clone" are met
--执行以下命令完成从库搭建
repmgr-h host01 -U repluser -d repmgr -f etc/repmgr.conf standby clone
注意：如果数据量很大，则在搭建从库时需要加上--fast-checkpoint参数，否则调用pg_basebackup备份的过程会非常慢。
---如果不是使用repmgr完成的流复制搭建，则配置连接串时，需要指定application_name
repmgr-f etc/repmgr.conf standby register
repmgr-f etc/repmgr.conf cluster show
至此，repmgr就搭建及配置完成。
--直接进行主从切换
repmgrstandby switchover -f etc/repmgr.conf --siblings-follow --dry-run --切换检查
repmgr-f /etc/repmgr.conf standby switchover
--主库down掉后，从库手动切换成主库
repmgr-f etc/repmgr.conf standby promote ---备节点
repmgr-f etc/repmgr.conf standby follow --如果有多个备节点，在其他备节点执行
--修改/etc/repmgr.conf
shared_preload_libraries = 'repmgr' monitor_interval_secs=2 connection_check_type=connection reconnect_attempts==6 reconnect_interval=4 failover=automatic promote_command='/usr/local/postgresql/bin/repmgr standby promote -f etc/repmgr.conf --log-to-file' follow_command='/usr/local/postgresql/bin/repmgr standby follow -f etc/repmgr.conf --log-to-file --upstream-node-id=%n' log_level='INFO' log_facility='STDERR' log_file='/data/pgdata/log/repmgr.log' log_status_interval=300 repmgrd_service_start_command = 'repmgrd --daemonize=true' repmgrd_service_stop_command = 'kill `cat data/pgdata/repmgrd.pid`' repmgrd_pid_file=' /pgdata/repmgrd.pid' |
--启动守护进程
repmgrdaemon start
--自动failover之后需手工将原主节点恢复成备节点，然后执行repmgr-f /etc/repmgr.conf standby register --force ，否则下次不会自动failover
keepalived在此架构中，只作为提供VIP的工具，不进行故障转移操作，所以配置相对简单，其配置如下：
/etc/keepalived/keepalived.conf
global_defs { router_id pg_ha # enable_script_security } vrrp_script checkpg { script "/etc/keepalived/scripts/checkpg.sh" interval 15 fall 3 rise 1 } vrrp_instance VI_pgusdp { state BACKUP interface ens160 virtual_router_id 152 priority 80 advert_int 1 nopreempt authentication { auth_type PASS auth_pass 234235 } track_script { checkpg } notify_master "/etc/keepalived/scripts/master.sh" notify_backup "/etc/keepalived/scripts/slave.sh" virtual_ipaddress { 10.**.**.**/24 } } |
/etc/keepalived/scripts/checkpg.sh
#!/bin/bash export PGDATABASE=postgres export PGPORT=5432 export PGUSER=postgres export PGBIN=/usr/local/postgresql/bin export PGDATA=/data/pgdata LOGFILE=/etc/keepalived/log/keepalived.log nc -w 3 localhost 5432 </dev/null a=`echo $?` if [ $a -eq 1 ] ;then exit 1 else SQL1='SELECT pg_is_in_recovery from pg_is_in_recovery();' db_role=`echo $SQL1 | ${PGBIN}/psql -d $PGDATABASE -U $PGUSER -h $PGDATA -At -w` if [ $db_role == 't' ];then exit 1 fi fi |
/etc/keepalived/scripts/master.sh
LOGFILE=/etc/keepalived/log/keepalived.log export PGDATABASE=postgres export PGPORT=5432 export PGUSER=postgres export PGBIN=/usr/local/postgresql/bin export PGDATA=/data/pgdata LOGFILE=/etc/keepalived/log/keepalived.log SQL1='SELECT pg_is_in_recovery from pg_is_in_recovery();' db_role=`echo $SQL1 | ${PGBIN}/psql -d $PGDATABASE -U $PGUSER -h $PGDATA -At -w` if [ $db_role == 't' ];then echo -e `date +"%F %T"` "the current database is standby DB! " >> $LOGFILE exit 1 else echo -e `date +"%F %T"` "the current database is master DB!" >> $LOGFILE fi |
/etc/keepalived/scripts/slave.sh
LOGFILE=/etc/keepalived/log/keepalived.log export PGDATABASE=postgres export PGPORT=5432 export PGUSER=postgres export PGBIN=/usr/local/postgresql/bin export PGDATA=/data/pgdata LOGFILE=/etc/keepalived/log/keepalived.log SQL1='SELECT pg_is_in_recovery from pg_is_in_recovery();' db_role=`echo $SQL1 | ${PGBIN}/psql -d $PGDATABASE -U $PGUSER -h $PGDATA -At -w` if [ $db_role == 't' ];then echo -e `date +"%F %T"` "the current database is standby DB! " >> $LOGFILE else echo -e `date +"%F %T"` "the current database is master DB!" >> $LOGFILE fi |
