Docker容器之Patroni篇

Posted 2022-01-21 PostgreSQLChina

作者：王志斌

上一篇文章向大家介绍了Repmgr的搭建过程，实现了自动切换，今天将向大家介绍，如何搭建容器下的Patroni集群环境，Patroni作为开箱即用PG高可用工具，越来越多的被各个厂商用于云环境下使用。

patroni基本架构如图所示：

etcd作为分布式注册中心、进行集群选主工作；vip-manager为主节点设置漂移IP；patroni负责引导集群的创建、运行和管理工作，并可以使用patronictl来进行终端访问。

具体流程：
1、首先启动etcd集群，本例中etcd数量为3个。
2、检测etcd集群状态健康后，启动patroni并竞争选主，其他跟随节点进行数据同步过程。
3、启动vip-manager，通过访问etcd集群中/$SERVIC{E}_{N}AME/$CLUSTER_NAME/leader键中的具体值，判断当前节点是否为主节点ip，如果是则为该节点设置vip，提供对外读写服务。
注：建议真实环境下将etcd部署到独立容器上，对外提供服务。

创建镜像

文件结构

其中Dockerfile为镜像主文件，docker服务通过该文件在本地仓库创建镜像；entrypoint.sh为容器入口文件，负责业务逻辑的处理；function为执行业务方法的入口文件，负责启动etcd，监控etcd集群状态、启动patroni和vip-manager；generatefile为整个容器生成对应的配置文件，包括etcd、patroni及vip-mananger。

目录结构大致如图所示：

注：数据库安装包和patroni安装包请自行构建。

DockerFile

FROM centos:7

MAINTAINER wangzhibin <wangzhibin>

ENV USER="postgresql" \\
    PASSWORD=123456 \\
    GROUP=postgresql 
	
RUN useradd $USER \\
       && chown -R $USER:$GROUP /home/$USER \\
       && yum -y update && yum install -y iptables sudo net-tools iproute  openssh-server openssh-clients which vim sudo crontabs
#安装etcd
COPY etcd/etcd /usr/sbin
COPY etcd/etcdctl /usr/sbin

#安装database
COPY lib/ /home/$USER/lib
COPY include/ /home/$USER/include
COPY share/ /home/$USER/share
COPY bin/ /home/$USER/bin/
COPY patroni/ /home/$USER/patroni

#安装vip-manager
COPY vip-manager/vip-manager /usr/sbin
#安装执行脚本
COPY runtime/ /home/$USER/runtime
COPY entrypoint.sh /sbin/entrypoint.sh

#设置环境变量
ENV LD_LIBRARY_PATH /home/$USER/lib
ENV PATH /home/$USER/bin:$PATH
ENV ETCDCTL_API=3

#安装Patroni
RUN yum -y install epel-release python-devel  && yum -y install python-pip \\
    && pip install /home/$USER/patroni/1/pip-20.3.3.tar.gz \\
    && pip install /home/$USER/patroni/1/psycopg2-2.8.6-cp27-cp27mu-linux_x86_64.whl \\
    && pip install --no-index --find-links=/home/$USER/patroni/2/ -r /home/$USER/patroni/2/requirements.txt \\
    && pip install /home/$USER/patroni/3/patroni-2.0.1-py2-none-any.whl

#修改执行权限
RUN chmod 755 /sbin/entrypoint.sh \\ 
&& mkdir /home/$USER/etcddata \\
&& chown -R $USER:$GROUP /home/$USER \\
&& echo 'root:root123456' | chpasswd \\
&& chmod 755 /sbin/etcd \\
&& chmod 755 /sbin/etcdctl \\
&& chmod 755 /sbin/vip-manager

#设置Sudo
RUN chmod 777 /etc/sudoers \\
       && sed -i '/## Allow root to run any commands anywhere/a '$USER' ALL=(ALL) NOPASSWD:ALL' /etc/sudoers \\
       && chmod 440 /etc/sudoers

#切换用户
USER $USER

#切换工作目录
WORKDIR /home/$USER

#启动入口程序
CMD ["/bin/bash", "/sbin/entrypoint.sh"]

entrypoint.sh

#!/bin/bash
set -e

# shellcheck source=runtime/functions
source "/home/$USER/runtime/function"

configure_patroni

function

#!/bin/bash

set -e
source /home/$USER/runtime/env-defaults
source /home/$USER/runtime/generatefile

PG_DATADIR=/home/$USER/pgdata
PG_BINDIR=/home/$USER/bin

configure_patroni()

    #生成配置文件
    generate_etcd_conf
    generate_patroni_conf
    generate_vip_conf
    #启动etcd
    etcdcount=$ETCD_COUNT
    count=0
    ip_temp=""
    array=($HOSTLIST//,/ )
    for host in $array[@]
    do
        ip_temp+="http://$host:2380,"
    done
    etcd --config-file=/home/$USER/etcd.yml >/home/$USER/etcddata/etcd.log 2>&1 &
    while [ $count -lt $etcdcount ]
    do
      line=(`etcdctl --endpoints=$ip_temp%? endpoint health -w json`)
      count=`echo $line | awk -F"\\"health\\":true" 'print NF-1'`
      echo "waiting etcd cluster"
      sleep 5
    done
    #启动patroni
    patroni /home/$USER/postgresql.yml >  /home/$USER/patroni/patroni.log 2>&1 &
    #启动vip-manager
    sudo vip-manager --config /home/$USER/vip.yml

generatefile

#!/bin/bash
set -e

HOSTNAME="`hostname`"
hostip=`ping $HOSTNAME -c 1 -w 1 | sed '1s/[^(]*(//;s/).*//;q'`

#generate etcd
generate_etcd_conf()

    echo "name : $HOSTNAME" >> /home/$USER/etcd.yml
    echo "data-dir: /home/$USER/etcddata" >> /home/$USER/etcd.yml
    echo "listen-client-urls: http://0.0.0.0:2379" >> /home/$USER/etcd.yml
    echo "advertise-client-urls: http://$hostip:2379" >> /home/$USER/etcd.yml
    echo "listen-peer-urls: http://0.0.0.0:2380" >> /home/$USER/etcd.yml
    echo "initial-advertise-peer-urls: http://$hostip:2380" >> /home/$USER/etcd.yml
    ip_temp="initial-cluster: "
    array=($HOSTLIST//,/ )  
    for host in $array[@]
    do
    	ip_temp+="$host=http://$host:2380," 
    done
    echo  $ip_temp%? >> /home/$USER/etcd.yml
    echo "initial-cluster-token: etcd-cluster-token" >> /home/$USER/etcd.yml
    echo "initial-cluster-state: new" >> /home/$USER/etcd.yml


#generate patroni
generate_patroni_conf()

  echo "scope: $CLUSTER_NAME" >> /home/$USER/postgresql.yml
  echo "namespace: /$SERVICE_NAME/ " >> /home/$USER/postgresql.yml
  echo "name: $HOSTNAME " >> /home/$USER/postgresql.yml
  echo "restapi: " >> /home/$USER/postgresql.yml 
  echo "  listen: $hostip:8008 " >> /home/$USER/postgresql.yml 
  echo "  connect_address: $hostip:8008 " >> /home/$USER/postgresql.yml
  echo "etcd: " >> /home/$USER/postgresql.yml
  echo "  host: $hostip:2379 " >> /home/$USER/postgresql.yml
  echo "  username: $ETCD_USER " >>  /home/$USER/postgresql.yml
  echo "  password: $ETCD_PASSWD " >> /home/$USER/postgresql.yml
  echo "bootstrap: " >> /home/$USER/postgresql.yml
  echo "  dcs: " >> /home/$USER/postgresql.yml
  echo "    ttl: 30 " >> /home/$USER/postgresql.yml
  echo "    loop_wait: 10  " >> /home/$USER/postgresql.yml
  echo "    retry_timeout: 10   " >> /home/$USER/postgresql.yml
  echo "    maximum_lag_on_failover: 1048576 " >> /home/$USER/postgresql.yml
  echo "    postgresql: " >>  /home/$USER/postgresql.yml
  echo "      use_pg_rewind: true  " >>  /home/$USER/postgresql.yml
  echo "      use_slots: true  " >>  /home/$USER/postgresql.yml
  echo "      parameters:  " >>  /home/$USER/postgresql.yml
  echo "  initdb:  " >>  /home/$USER/postgresql.yml
  echo "  - encoding: UTF8  " >>  /home/$USER/postgresql.yml
  echo "  - data-checksums  " >>  /home/$USER/postgresql.yml
  echo "  pg_hba:   " >>  /home/$USER/postgresql.yml
  echo "  - host replication $USER 0.0.0.0/0 md5  " >>  /home/$USER/postgresql.yml
  echo "  - host all all 0.0.0.0/0 md5  " >>  /home/$USER/postgresql.yml
  echo "postgresql:  " >>  /home/$USER/postgresql.yml
  echo "  listen: 0.0.0.0:5432  " >>  /home/$USER/postgresql.yml
  echo "  connect_address: $hostip:5432  " >>  /home/$USER/postgresql.yml
  echo "  data_dir: $PG_DATADIR  " >>  /home/$USER/postgresql.yml
  echo "  bin_dir: $PG_BINDIR  " >>  /home/$USER/postgresql.yml
  echo "  pgpass: /tmp/pgpass  " >>  /home/$USER/postgresql.yml
  echo "  authentication:  " >>  /home/$USER/postgresql.yml
  echo "    replication:  " >>  /home/$USER/postgresql.yml
  echo "      username: $USER  " >>  /home/$USER/postgresql.yml
  echo "      password: $PASSWD  " >>  /home/$USER/postgresql.yml
  echo "    superuser:  " >>  /home/$USER/postgresql.yml
  echo "      username: $USER  " >>  /home/$USER/postgresql.yml
  echo "      password: $PASSWD  " >>  /home/$USER/postgresql.yml
  echo "    rewind:  " >>  /home/$USER/postgresql.yml
  echo "      username: $USER  " >>  /home/$USER/postgresql.yml
  echo "      password: $PASSWD  " >>  /home/$USER/postgresql.yml
  echo "  parameters:  " >>  /home/$USER/postgresql.yml
  echo "    unix_socket_directories: '.'  " >>  /home/$USER/postgresql.yml
  echo "    wal_level: hot_standby  " >>  /home/$USER/postgresql.yml
  echo "    max_wal_senders: 10  " >>  /home/$USER/postgresql.yml
  echo "    max_replication_slots: 10  " >>  /home/$USER/postgresql.yml
  echo "tags:  " >>  /home/$USER/postgresql.yml
  echo "    nofailover: false  " >>  /home/$USER/postgresql.yml
  echo "    noloadbalance: false  " >>  /home/$USER/postgresql.yml
  echo "    clonefrom: false  " >>  /home/$USER/postgresql.yml
  echo "    nosync: false  " >>  /home/$USER/postgresql.yml

#........ 省略部分内容

构建镜像

docker build -t patroni .

运行镜像

运行容器节点1：
docker run --privileged --name patroni1 -itd --hostname patroni1 --net my_net3 --restart always --env ‘CLUSTER_NAME=patronicluster’ --env ‘SERVICE_NAME=service’ --env ‘ETCD_USER=etcduser’ --env ‘ETCD_PASSWD=etcdpasswd’ --env ‘PASSWD=zalando’ --env ‘HOSTLIST=patroni1,patroni2,patroni3’ --env ‘VIP=172.22.1.88’ --env ‘NET_DEVICE=eth0’ --env ‘ETCD_COUNT=3’ patroni
运行容器节点2：
docker run --privileged --name patroni2 -itd --hostname patroni2 --net my_net3 --restart always --env ‘CLUSTER_NAME=patronicluster’ --env ‘SERVICE_NAME=service’ --env ‘ETCD_USER=etcduser’ --env ‘ETCD_PASSWD=etcdpasswd’ --env ‘PASSWD=zalando’ --env ‘HOSTLIST=patroni1,patroni2,patroni3’ --env ‘VIP=172.22.1.88’ --env ‘NET_DEVICE=eth0’ --env ‘ETCD_COUNT=3’ patroni
运行容器节点3：
docker run --privileged --name patroni3 -itd --hostname patroni3 --net my_net3 --restart always --env ‘CLUSTER_NAME=patronicluster’ --env ‘SERVICE_NAME=service’ --env ‘ETCD_USER=etcduser’ --env ‘ETCD_PASSWD=etcdpasswd’ --env ‘PASSWD=zalando’ --env ‘HOSTLIST=patroni1,patroni2,patroni3’ --env ‘VIP=172.22.1.88’ --env ‘NET_DEVICE=eth0’ --env ‘ETCD_COUNT=3’ patroni

总结

本操作过程仅限于试验环境，为了演示etcd+patroni+vipmanager整体的容器化搭建。在真实环境下，etcd应该部署在不同容器下，形成独立的分布式集群，并且PG的存储应该映射到本地磁盘或网络磁盘，另外容器集群的搭建尽量使用编排工具，例如docker-compose、docker-warm或者Kubernetes等。

附图

etcd集群状态如图：

patroni集群状态如图：

vip-manager状态如图：