作者:王志斌,曾获得中国PostgreSQL数据库管理工程师(PGCE),是PostgreSQL官方认证讲师,盘古云课堂特邀金牌讲师。 最后再来说一下关于Pgbouncer的部署形式,包括单应用场景、多应用场景、集群场景还有多实例场景,这些方式都是依据不同的业务场景,没有孰优孰劣,符合的才是对的。其中单应用和多应用场景来源于官方。 单应用场景:

图 9 单应用多连接场景结构图 单应用场景主要具体为短连接较多的场景,频繁进行数据库的连接操作,但操作时间较短,均为短连接,所以将pgbouncer于应用服务器部署在同一台服务器,减少应用服务器和pgbouncer之间的开销。 配置文件
[databases] test1 = test = [pgbouncer] listen_port = 6688 listen_addr = 192.168.165.3 auth_type = md5 auth_file = /home/postgres/pgbouncer/bin/userlist.txt logfile = /home/postgres/pgbouncer/pgbouncer1.log pidfile =/home/postgres/pgbouncer/pgbouncer1.pid unix_socket_dir = /tmp ;;unix_socket_mode = 0777 admin_users = wzb stats_users = wzb pool_mode = session max_client_conn=1000 default_pool_size=30
导出数据库中用户名及密码到userslist.txt userslist.txt,格式为用户名 密码
"testuser" "md54d15115d8bebd3188c1ae09c4a9848af" "testuser1" "md5f8386abbae413786661ee5a5cfb5593c" "wzb" "md53d57c4bc9a647385e6916efd0b44db46"
启动Pgbouncer pgbouncer -d pgbouncer.ini 客户端连接方式 psql -dtest1 -Utestuser1 -p6688 多应用场景:

图 10多应用场景结构图 多应用场景,一般指多个应用服务器连接数据库,因此可以选择将pgbouncer与数据库服务部署在同一台服务器上,减少pgbouncer和数据库之间的开销。 配置PgBouncer.ini文件
[databases] a1 = host=127.0.0.1 port=5432 dbname=test a2 = host=127.0.0.1 port=5432 dbname=test1 [pgbouncer] listen_port = 6688 listen_addr = * auth_type = md5 auth_file = /home/postgres/pgbouncer/bin/userlist.txt logfile = /home/postgres/pgbouncer/pgbouncer.log pidfile =/home/postgres/pgbouncer/pgbouncer.pid admin_users = wzb stats_users = wzb pool_mode = session max_client_conn=1000 default_pool_size=30
导出数据库中用户名及密码到userslist.txt userslist.txt,格式为用户名 密码
"testuser" "md54d15115d8bebd3188c1ae09c4a9848af" "testuser1" "md5f8386abbae413786661ee5a5cfb5593c" "wzb" "md53d57c4bc9a647385e6916efd0b44db46"
启动Pgbouncer pgbouncer -d pgbouncer.ini 连接后端数据库 $ psql -p 6688 -U testuser a1 $ psql -p 6688 -U testuser1 a2 连接pgbouncer数据库 psql -p 6688 pgbouncer -U wzb pgbouncer=# show help; NOTICE: Console usage DETAIL: SHOW HELP|CONFIG|DATABASES|POOLS|CLIENTS|SERVERS|USERS|VERSION SHOW FDS|SOCKETS|ACTIVE_SOCKETS|LISTS|MEM SHOW DNS_HOSTS|DNS_ZONES SHOW STATS|STATS_TOTALS|STATS_AVERAGES|TOTALS SET key = arg RELOAD PAUSE [<db>] RESUME [<db>] DISABLE <db> ENABLE <db> RECONNECT [<db>] KILL <db> SUSPEND SHUTDOWN SHOW pgbouncer=# show clients;
type| C user| pgbouncer database| pgbouncer state| active addr| unix port| 6432 local_addr| unix local_port| 6432 connect_time| 2020-10-09 20:41:32 CST request_time| 2020-10-09 20:41:32 CST wait| 5 wait_us| 483185 close_needed| 0 ptr| 0x9ec340 link| remote_pid| 23567 tls |
pgbouncer=# show pools;
database| pgbouncer user| pgbouncer cl_active| 1 cl_waiting| 0 sv_active|0 sv_idle|0 sv_used|0 sv_tested|0 sv_login|0 maxwait|0 maxwait_us|0 pool_mode| transaction
集群场景(读写分离): 读写分离场景下pgbouncer的配置与前面配置基本一致,主要区别于要针对读和写进行分别部署pgbouncer,因为pgbouncer本身只是数据库连接池,不具备负载均衡,或高可用,IP漂移等特性,需要结合其他成熟产品进行组合使用。 多实例场景:

图 11 多实例场景结构图 多实例场景主要利用linux系统端口重用技术,这个特性依靠Linux内核上的支持(Linux3.6以上版本),并结合pgbouncer自身支持(设置so_reuseport=1)结合起来形成多实例场景下的pgbouncer使用,可以认为是pgbouncer的高可靠或者高可用,在某一个实例进程故障的情况下,其他实例集成仍然可以处理来自外部的数据库连接请求。从操作系统层面来看,属于多进程共享同一个端口。 实例配置1
[databases] a2 = host=127.0.0.1 port=5432 dbname=test1 pool_size=50 ;;a1 = host=127.0.0.1 port=5432 dbname=test pool_size=30 [pgbouncer] listen_port = 6688 listen_addr = 192.168.165.3 auth_type = md5 auth_file = /home/postgres/pgbouncer/bin/userlist.txt logfile = /home/postgres/pgbouncer/pgbouncer1.log pidfile =/home/postgres/pgbouncer/pgbouncer1.pid unix_socket_dir = /tmp/pg1 #unix_socket_mode = 0777 admin_users = wzb stats_users = wzb pool_mode = session max_client_conn=1000 default_pool_size=30 so_reuseport = 1
实例配置2
[databases] a2 = host=127.0.0.1 port=5432 dbname=test1 pool_size=50 ;;a1 = host=127.0.0.1 port=5432 dbname=test pool_size=30 [pgbouncer] listen_port = 6688 listen_addr = 192.168.165.3 auth_type = md5 auth_file = /home/postgres/pgbouncer/bin/userlist.txt logfile = /home/postgres/pgbouncer/pgbouncer2.log pidfile =/home/postgres/pgbouncer/pgbouncer2.pid unix_socket_dir = /tmp/pg2 #unix_socket_mode = 0777 admin_users = wzb stats_users = wzb pool_mode = session max_client_conn=1000 default_pool_size=30 so_reuseport = 1
导出数据库中用户名及密码到userslist.txt userslist.txt,格式为用户名 密码
"testuser" "md54d15115d8bebd3188c1ae09c4a9848af" "testuser1" "md5f8386abbae413786661ee5a5cfb5593c" "wzb" "md53d57c4bc9a647385e6916efd0b44db46"
启动多实例 ./pgbouncer pgbouncer.ini ./pgbouncer pgbouncer1.ini 参考 [1]Pgbouncer官网 [2]PgBouncer Configuration [3]Tuning PostgreSQL for sysbench-tpcc [4]understanding-user-management-in-pgbouncer [5]performance-best-practices-for-using-azure-database-for-postgresql-connection-pooling [6]guide-using-pgbouncer [7]azure-database-for-postgresql/connection-handling-best-practice-with-postgresql [8]steps-to-install-and-setup-pgbouncer-connection-pooling-proxy [9]pg-phriday-securing-pgbouncer 了解更多PostgreSQL热点资讯、新闻动态、精彩活动,请访问中国PostgreSQL官方网站: 解决更多PostgreSQL相关知识、技术、工作问题,请访问中国PostgreSQL官方问答社区: 下载更多PostgreSQL相关资料、工具、插件问题,请访问中国PostgreSQL官方下载网站:
