關於Consul

Consul 是一個支援多資料中心分散式高可用的服務發現和配置共享的服務軟體.
Consul 由 HashiCorp公司用Go語言開發, 基於Mozilla Public License 2.0的協議進行開源.
Consul 支援健康檢查,並允許 HTTP 和 DNS 協議呼叫 API 儲存鍵值對.
命令列超級好用的虛擬機器管理軟體 vgrant 也是 HashiCorp 公司開發的產品.
一致性協議採用 Raft 演算法,用來保證服務的高可用. 使用 GOSSIP 協議管理成員和廣播訊息, 並且支援 ACL 訪問控制.

關於官方consul映象

關於官方consul映象，三點說明:
+ 想對於其他版本映象，其stars和pull數量是最多的, pull達到10M+;
+ 官方文件可讀性相對不夠好, 有些繞;
+ 目前官方映象更新進度基本與consul最新版本保持一致。

多機部署consul:0.8.5版本操作參考

部分引數說明

• –net=host docker引數, 使得docker容器越過了net namespace的隔離，免去手動指定埠對映的步驟
• -server consul支援以server或client的模式執行, server是服務發現模組的核心, client主要用於轉發請求
• -advertise 將本機私有IP傳遞到consul
• -bootstrap-expect 指定consul將等待幾個節點連通，成為一個完整的叢集
• -retry-join 指定要加入的consul節點地址，失敗會重試, 可多次指定不同的地址
• -client consul繫結在哪個client地址上，這個地址提供HTTP、DNS、RPC等服務，預設是127.0.0.1
• -bind 該地址用來在叢集內部的通訊，叢集內的所有節點到地址都必須是可達的，預設是0.0.0.0
• allow_stale 設定為true, 表明可以從consul叢集的任一server節點獲取dns資訊, false則表明每次請求都會經過consul server leader

叢集的多機部署參考

• 多中心部署結構圖

  
  生產環境下, 一般一個宿主host一個consul節點；
  server節點建議一個數據中心部署3-5個, client節點可部署任意節點。

• 啟動consul server節點, 在對應的機器上root許可權下執行以下指令碼(需修改節點列表: nodes=(…) ):

#!/bin/bash
 

conf_dir=/opt/consul/conf
mkdir -p ${conf_dir}

nodes=(
192.168.1.10
192.168.1.11
192.168.1.12
)
consul_ver=0.8.5
retry_interval=15s
contain_svr_name=consul_server

privIP=$(/sbin/ifconfig eth0 | sed -n 's/.*inet \(addr:\)\?\([0-9.]\{7,15\}\) .*/\2/p')
if [[ !("${nodes[@]}" =~ $privIP) ]];
then
    echo -e "Current node:${privIP} not in configured server nodes.\n"
    exit
fi

svr_runing=$(docker ps -a | grep "${contain_svr_name}" | egrep "Up [About]|[0-9]{1,}")
if [[ ${svr_runing} != "" ]];
then
    echo -e "Current container of consul server has been running.\n"
    exit
else
    svr_exists=$(docker ps -a | grep "${contain_svr_name}")
    if [[ ${svr_exists} != "" ]];
    then
        echo -e "Now try to start the container as it stopped...\n"
        docker start ${svr_exists}
        sleep 2
        docker ps -a grep "${contain_svr_name}"
        exit
    fi
fi

echo -e "To start a new container for consul...\n"
echo -e "To initialize configuration...\n"

nodels=""
for host in ${nodes[*]}
do
    if [[ $nodels != "" ]];
    then
        nodels=$nodels,
    fi
    nodels=$nodels"\"$host\""
done

config="{\n
\"datacenter\": \"dctest\",\n
\"retry_join\": [${nodels}],\n
\"retry_interval\": \"${retry_interval}\",\n
\"rejoin_after_leave\": true,\n
\"start_join\": [${nodels}],\n
\"bootstrap_expect\": 3,\n
\"server\": true,\n
\"ui\": true,\n
\"dns_config\": {\"allow_stale\": true, \"max_stale\": \"5s\"},\n
\"node_name\": \"$HOSTNAME\"\n
}\n"

echo $config
echo -e ${config} > ${conf_dir}/server.json
echo -e ${config}

docker run -d -v ${conf_dir}:${conf_dir} \
    --name ${contain_svr_name} \
    --net=host consul:${consul_ver} agent \
    -config-dir=${conf_dir} \
    -client=0.0.0.0 \
    -bind=${privIP} \
    -advertise=${privIP}

sleep 2

svr_runing=$(docker ps -a | grep "${contain_svr_name}" | egrep "Up [About]|[0-9]{1,}")
if [[ ${svr_runing} == "" ]];
then
    echo -e "\nError: docker-consul failed to start...\n"
    exit
fi
echo -e "\nOK: docker-consul has started as background server.\n"

• 啟動consul client節點(任意數量或不部署), 在需要部署的機器上執行(需修改svrnodes列表):

#!/bin/bash

conf_dir=/opt/consul/conf
mkdir -p ${conf_dir}

svrnodes=(
192.168.1.10
192.168.1.11
192.168.1.12
)
privIP=$(/sbin/ifconfig eth0 | sed -n 's/.*inet \(addr:\)\?\([0-9.]\{7,15\}\) .*/\2/p')
consul_ver=0.8.5
retry_interval=15s
contain_cli_name=consul_client

if [[ "${svrnodes[@]}" =~ $privIP ]];
then
    echo -e "Current node:${privIP} is configured for server consul, not to run client mode.\n"
    exit
fi

svr_runing=$(docker ps -a | grep "${contain_cli_name}" | egrep "Up [About]|[0-9]{1,}")
if [[ ${svr_runing} != "" ]];
then
    echo -e "Current container of consul client has been running.\n"
    exit
else
    svr_exists=$(docker ps -a | grep "${contain_cli_name}")
    if [[ ${svr_exists} != "" ]];
    then
        echo -e "Now try to start the container as it stopped...\n"
        docker start ${svr_exists}
        sleep 2
        docker ps -a grep "${contain_cli_name}"
        exit
    fi
fi

echo -e "To start a new container for consul...\n"
echo -e "To initialize configuration...\n"

nodels=""
for host in ${svrnodes[*]}
do
    if [[ $nodels != "" ]];
    then
        nodels=$nodels,
    fi
    nodels=$nodels"\"$host\""
done

config="{\n
\"retry_join\": [${nodels}],\n
\"retry_interval\": \"${retry_interval}\",\n
\"rejoin_after_leave\": true,\n
\"start_join\": [${nodels}],\n
\"server\": false,\n
\"ui\": true,\n
\"node_name\": \"$HOSTNAME\"\n
}\n"

echo $config
echo -e ${config} > ${conf_dir}/client.json
echo -e ${config}

docker run -d -v ${conf_dir}:${conf_dir} \
    --name ${contain_cli_name} \
    --net=host consul:${consul_ver} agent \
    -config-dir=${conf_dir} \
    -client=0.0.0.0 \
    -advertise=${privIP}

sleep 2

svr_runing=$(docker ps -a | grep "${contain_cli_name}" | egrep "Up [About]|[0-9]{1,}")
if [[ ${svr_runing} == "" ]];
then
    echo -e "\nError: docker-consul client node failed to start...\n"
    exit
fi
echo -e "\nOK: docker-consul has started as a client node.\n"

• 檢查叢集的狀態

  通過webUI地址: http://<其中一個節點ip>:8500/ui/可檢視叢集狀態，正常情況下會顯示3 passing。
  通過docker命令列:
  

[email protected]:~# docker exec -it consul_server consul members
Node                          Address             Status  Type    Build  Protocol  DC
hostname1  192.168.1.10:8301  alive   server  0.8.5  2         dctest
hostname2  192.168.1.11:8301  alive   server  0.8.5  2         dctest
hostname3  192.168.1.12:8301  alive   server  0.8.5  2         dctest
hostname4  192.168.1.13:8301  alive   client  0.8.5  2         dctest