部署 PostgreSQL 流复制

按实际环境修改
- docker-compose.yml
- pgsql1/conf.d/*.cnf
- pgsql2/conf.d/*.cnf

创建目录

grep '\<source:' docker-compose.yml | cut -d: -f2 | xargs mkdir -p
chmod 0750 pgsql{1..2}/db

启动
```
docker-compose up -d
```
查看日志，直至出现 "PGSQL is ready for connections."
```
docker-compose logs pgsql1
docker-compose logs pgsql2
```

docker exec -u postgres pgsql1 \
    psql -c "create table t1(id int, name varchar)"
docker exec -u postgres pgsql1 \
    psql -c "insert into t1 values (1, 'aa'),(2, 'bb')"

docker exec -u postgres pgsql1 \
    psql -c "create user replicator login replication password 'Replicator_1234'"
docker exec -u postgres pgsql1 \
    echo 'host replication replicator pgsql2 trust' >> pgsql1/db/pg_hba.conf
#docker exec -u postgres pgsql1 pg_ctl reload

增加主库配置

echo '
wal_level = hot_standby
max_wal_senders = 8
wal_keep_size = 512
wal_sender_timeout = 60s
max_connections = 100
' > pgsql1/conf.d/replication.conf

重启 pgsql1 数据库
```
docker-compose restart pgsql1
```

docker exec -u postgres pgsql2 mkdir -p /var/log/pgsql/db_pg1
docker exec -u postgres pgsql2 \
    pg_basebackup \
        -h pgsql1 \
        -D /var/log/pgsql/db_pg1 \
        -p 5432 \
        -U replicator \
        -Fp -Xs -P -c fast

停止并删除 pgsql2 容器，替换其数据目录，注意数据目录权限

docker-compose stop pgsql2
docker-compose rm -f pgsql2
rm -rf pgsql2/db
mv pgsql2/log/db_pg1 pgsql2/db
chmod 0750 pgsql2/db

增加从库配置

echo '
primary_conninfo = 'host=pgsql1 port=5432 user=replicator'
recovery_target_timeline = latest
hot_standby = on
max_standby_streaming_delay = 30s
wal_receiver_status_interval = 10s
hot_standby_feedback = on
max_connections = 200
max_worker_processes = 20
' > pgsql2/conf.d/replication.conf

启动 pgsql2 数据库
```
docker-compose up -d pgsql2
```

docker exec -u postgres pgsql1 \
    psql -c "select
        client_addr,
        usename,
        backend_start,
        application_name,
        sync_state,
        sync_priority
        FROM pg_stat_replication;"
# 在 pgsql1 里创建数据库、表或更改数据，在 pgsql2 里可以看到

查看状态

docker exec -u postgres pgsql1 pg_controldata -D /var/lib/pgsql | grep cluster
# 主库(pgsql1)输出: in production
docker exec -u postgres pgsql2 pg_controldata -D /var/lib/pgsql | grep cluster
# 从库(pgsql2)输出: in archive recovery

从库提升为主库

docker exec -u postgres pgsql2 pg_ctl promote

3.1 KiB Raw Blame History

部署 PostgreSQL 流复制

3.1 KiB

Raw Blame History