分布式服务注册发现与统一配置管理之 Consul
共 12990字,需浏览 26分钟
·
2020-09-02 16:23
Hello 大家好,我是阿粉,前面的文章给大家介绍过 Nacos,用于服务注册发现和管理配置的开源组件,今天给大家分享另一个组件 Consul 也有相应的功能,我们一起来看一下吧!
背景
目前分布式系统架构已经基本普及,很多项目都是基于分布式架构的,以往的单机模式基本已经不适应当下互联网行业的发展。随着分布式项目的普及,项目服务实例数目的增加,服务的注册与发现功能就成了一项必不可少的架构。服务的注册与发现的功能,有很多开源方案。包括早期的zookeeper,百度的disconf,阿里的diamond,基于Go语言的ETCD,Spring集成的Eureka,以及前文提到的 Nacos 还有本文的主角Consul。这里不对上面提到的进行比较,本文仅介绍Consul,详细的对比,说明网上有很多资料,可以参考,例如:服务发现比较:Consul vs Zookeeper vs Etcd vs Eureka说到服务的注册与发现主要是下面两个主要功能:
服务注册与发现 配置中心即分布式项目统一配置管理
Consul 简介
Consul采用Go语言开发 Consul内置服务注册与发现框架、分布一致性协议实现、健康检查、Key/Value存储、多数据中心方案;不依赖其他工具,安装包包含一个可执行文件 支持DNS、HTTP协议接口 自带web-ui Consul是 HashiCorp 公司推出的开源工具,用于实现分布式系统的服务发现与配置。 Consul支持两种服务注册的方式,一种是通过Consul的服务注册HTTP API,由服务自身在启动后调用API注册自己,另外一种则是通过在配置文件中定义服务的方式进行注册。Consul文档中建议使用后面一种方式来做服务 配置和服务注册。
Consul 服务端配置使用
下载相应版本解压,并将可执行文件复制到/usr/local/consul目录下
创建一个service的配置文件
silence$ sudo mkdir /etc/consul.d
silence$ echo '{"service":{"name": "web", "tags": ["rails"], "port": 80}}' | sudo tee /etc/consul.d/web.json
启动代理
local/consul/consul agent -dev -node consul_01 -config-dir=/etc/consul.d/ -ui /usr/
-dev 参数代表本地测试环境启动;-node 参数表示自定义集群名称;-config-drir 参数表示services的注册配置文件目录,即上面创建的文件夹-ui 启动自带的web-ui管理页面
集群成员查询方式
silence-pro:~ silence$ /usr/local/consul/consul members
HTTP协议数据查询
silence-pro:~ silence$ curl http://127.0.0.1:8500/v1/catalog/service/web
[
{
: ,
: ,
: ,
: ,
: {
: ,
:
},
: {
:
},
: ,
: ,
: [
],
"ServiceAddress": "",
"ServicePort": 80,
"ServiceEnableTagOverride": false,
"CreateIndex": 6,
"ModifyIndex": 6
}
]
silence-pro:~ silence$
web-ui管理
Consul Web UI
Consul的web-ui可以用来进行服务状态的查看,集群节点的检查,访问列表的控制以及KV存储系统的设置,相对于Eureka和ETCD,Consul的web-ui要好用的多。(Eureka和ETCD将在下一篇文章中简单介绍。)
7. KV存储的数据导入和导出
silence-pro:consul silence$ ./consul kv import .json
silence-pro:consul silence$ ./consul kv export redis/
temp.json文件内容格式如下,一般是管理页面配置后先导出保存文件,以后需要再导入该文件
[
{
"key": "redis/config/password",
"flags": 0,
"value": "MTIzNDU2"
},
{
"key": "redis/config/username",
"flags": 0,
"value": "U2lsZW5jZQ=="
},
{
"key": "redis/zk/",
"flags": 0,
"value": ""
},
{
"key": "redis/zk/password",
"flags": 0,
"value": "NDU0NjU="
},
{
"key": "redis/zk/username",
"flags": 0,
"value": "ZGZhZHNm"
}
]
Consul的KV存储系统是一种类似zk的树形节点结构,用来存储相关key/value键值对信息的,我们可以使用KV存储系统来实现上面提到的配置中心,将统一的配置信息保存在KV存储系统里面,方便各个实例获取并使用同一配置。而且更改配置后各个服务可以自动拉取最新配置,不需要重启服务。
Consul Java 客户端使用
maven pom依赖增加,版本可自由更换
<dependency>
<groupId>com.orbitz.consulgroupId>
<artifactId>consul-clientartifactId>
<version>0.12.3version>
dependency>
<dependency>
<groupId>org.springframework.bootgroupId>
<artifactId>spring-boot-starter-actuatorartifactId>
dependency>
Consul 基本工具类,根据需要相应扩展
package com.coocaa.consul.consul.demo;
import com.google.common.base.Optional;
import com.google.common.net.HostAndPort;
import com.orbitz.consul.*;
import com.orbitz.consul.model.agent.ImmutableRegCheck;
import com.orbitz.consul.model.agent.ImmutableRegistration;
import com.orbitz.consul.model.health.ServiceHealth;
import java.net.MalformedURLException;
import java.net.URI;
import java.util.List;
public class ConsulUtil {
private static Consul consul = Consul.builder().withHostAndPort(HostAndPort.fromString("127.0.0.1:8500")).build();
/**
* 服务注册
*/
public static void serviceRegister() {
AgentClient agent = consul.agentClient();
try {
/**
* 注意该注册接口:
* 需要提供一个健康检查的服务URL,以及每隔多长时间访问一下该服务(这里是3s)
*/
agent.register(8080, URI.create("http://localhost:8080/health").toURL(), 3, "tomcat", "tomcatID", "dev");
} catch (MalformedURLException e) {
e.printStackTrace();
}
}
/**
* 服务获取
*
* @param serviceName
*/
public static void findHealthyService(String serviceName) {
HealthClient healthClient = consul.healthClient();
List
serviceHealthList = healthClient.getHealthyServiceInstances(serviceName).getResponse(); serviceHealthList.forEach((response) -> {
System.out.println(response);
});
}
/**
* 存储KV
*/
public static void storeKV(String key, String value) {
KeyValueClient kvClient = consul.keyValueClient();
kvClient.putValue(key, value);
}
/**
* 根据key获取value
*/
public static String getKV(String key) {
KeyValueClient kvClient = consul.keyValueClient();
Optional<String> value = kvClient.getValueAsString(key);
if (value.isPresent()) {
return value.get();
}
return "";
}
/**
* 找出一致性的节点(应该是同一个DC中的所有server节点)
*/
public static List<String> findRaftPeers() {
StatusClient statusClient = consul.statusClient();
return statusClient.getPeers();
}
/**
* 获取leader
*/
public static String findRaftLeader() {
StatusClient statusClient = consul.statusClient();
return statusClient.getLeader();
}
public static void main(String[] args) {
AgentClient agentClient = consul.agentClient();
agentClient.deregister("tomcatID");
}
}
temp.json 和 ConsulUtil.java 文件以及上传到 GitHub 资源库,回复【源码仓库】获取代码地址。
通过上面的基本工具类可以实现服务的注册和KV数据的获取与存储功能
Consul集群搭建
参考地址 Consul集群搭建 http://ju.outofmemory.cn/entry/189899
三台主机Consul下载安装,我这里没有物理主机,所以通过三台虚拟机来实现。虚拟机IP分192.168.231.145,192.168.231.146,192.168.231.147
将145和146两台主机作为Server模式启动,147作为Client模式启动,Server和Client只是针对Consul集群来说的,跟服务没有任何关系!
Server模式启动145,节点名称设为n1,数据中心统一用dc1
[root@centos145 consul]# ./consul agent -server -bootstrap-expect 2 -data-dir /tmp/consul -node=n1 -bind=192.168.231.145 -datacenter=dc1
bootstrap_expect = 2: A cluster with 2 servers will provide no failure tolerance. See https://www.consul.io/docs/internals/consensus.html#deployment-table
bootstrap_expect > 0: expecting 2 servers
==> Starting Consul agent...
==> Consul agent running!
Version: 'v1.0.1'
Node ID: '6cc74ff7-7026-cbaa-5451-61f02114cd25'
Node name: 'n1'
Datacenter: 'dc1' (Segment: '
' )Server: true (Bootstrap: false)
Client Addr: [127.0.0.1] (HTTP: 8500, HTTPS: -1, DNS: 8600)
Cluster Addr: 192.168.231.145 (LAN: 8301, WAN: 8302)
Encrypt: Gossip: false, TLS-Outgoing: false, TLS-Incoming: false
==> Log data will now stream in as it occurs:
2017/12/06 23:26:21 [INFO] raft: Initial configuration (index=0): []
2017/12/06 23:26:21 [INFO] serf: EventMemberJoin: n1.dc1 192.168.231.145
2017/12/06 23:26:21 [INFO] serf: EventMemberJoin: n1 192.168.231.145
2017/12/06 23:26:21 [INFO] agent: Started DNS server 127.0.0.1:8600 (udp)
2017/12/06 23:26:21 [INFO] raft: Node at 192.168.231.145:8300 [Follower] entering Follower state (Leader: "")
2017/12/06 23:26:21 [INFO] consul: Adding LAN server n1 (Addr: tcp/192.168.231.145:8300) (DC: dc1)
2017/12/06 23:26:21 [INFO] consul: Handled member-join event for server "n1.dc1" in area "wan"
2017/12/06 23:26:21 [INFO] agent: Started DNS server 127.0.0.1:8600 (tcp)
2017/12/06 23:26:21 [INFO] agent: Started HTTP server on 127.0.0.1:8500 (tcp)
2017/12/06 23:26:21 [INFO] agent: started state syncer
2017/12/06 23:26:28 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:26:30 [WARN] raft: no known peers, aborting election
2017/12/06 23:26:49 [ERR] agent: Coordinate update error: No cluster leader
2017/12/06 23:26:54 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:27:24 [ERR] agent: Coordinate update error: No cluster leader
2017/12/06 23:27:27 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:27:56 [ERR] agent: Coordinate update error: No cluster leader
2017/12/06 23:28:02 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:28:27 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:28:33 [ERR] agent: Coordinate update error: No cluster leader
目前只启动了145,所以还没有集群
Server模式启动146,节点名称用n2,并在n2上启用了web-ui管理页面功能
[root@centos146 consul]# ./consul agent -server -bootstrap-expect 2 -data-dir /tmp/consul -node=n2 -bind=192.168.231.146 -datacenter=dc1 -ui
bootstrap_expect = 2: A cluster with 2 servers will provide no failure tolerance. See https://www.consul.io/docs/internals/consensus.html#deployment-table
bootstrap_expect > 0: expecting 2 servers
==> Starting Consul agent...
==> Consul agent running!
Version: 'v1.0.1'
Node ID: 'eb083280-c403-668f-e193-60805c7c856a'
Node name: 'n2'
Datacenter: 'dc1' (Segment: '
' )Server: true (Bootstrap: false)
Client Addr: [127.0.0.1] (HTTP: 8500, HTTPS: -1, DNS: 8600)
Cluster Addr: 192.168.231.146 (LAN: 8301, WAN: 8302)
Encrypt: Gossip: false, TLS-Outgoing: false, TLS-Incoming: false
==> Log data will now stream in as it occurs:
2017/12/06 23:28:30 [INFO] raft: Initial configuration (index=0): []
2017/12/06 23:28:30 [INFO] serf: EventMemberJoin: n2.dc1 192.168.231.146
2017/12/06 23:28:31 [INFO] serf: EventMemberJoin: n2 192.168.231.146
2017/12/06 23:28:31 [INFO] raft: Node at 192.168.231.146:8300 [Follower] entering Follower state (Leader: "")
2017/12/06 23:28:31 [INFO] consul: Adding LAN server n2 (Addr: tcp/192.168.231.146:8300) (DC: dc1)
2017/12/06 23:28:31 [INFO] consul: Handled member-join event for server "n2.dc1" in area "wan"
2017/12/06 23:28:31 [INFO] agent: Started DNS server 127.0.0.1:8600 (tcp)
2017/12/06 23:28:31 [INFO] agent: Started DNS server 127.0.0.1:8600 (udp)
2017/12/06 23:28:31 [INFO] agent: Started HTTP server on 127.0.0.1:8500 (tcp)
2017/12/06 23:28:31 [INFO] agent: started state syncer
2017/12/06 23:28:38 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:28:39 [WARN] raft: no known peers, aborting election
2017/12/06 23:28:57 [ERR] agent: Coordinate update error: No cluster leader
2017/12/06 23:29:11 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:29:30 [ERR] agent: Coordinate update error: No cluster leader
2017/12/06 23:29:38 [ERR] agent: failed to sync remote state: No cluster leader
2017/12/06 23:29:57 [ERR] agent: Coordinate update error: No cluster leader
同样没有集群发现,此时n1和n2都启动起来,但是互相并不知道集群的存在!
将n1节点加入n2
[join 192.168.231.146 ]$ ./consul
此时n1和n2都打印发现了集群的日志信息
这个时候n1和n2两个节点已经是一个集群里面的Server模式的节点了
Client模式启动147
[ ]
==> Starting Consul agent...
==> Consul agent running!
Version: 'v1.0.1'
Node ID: 'be7132c3-643e-e5a2-9c34-cad99063a30e'
Node name: 'n3'
Datacenter: 'dc1' (Segment: '')
Server: false (Bootstrap: false)
Client Addr: [127.0.0.1] (HTTP: 8500, HTTPS: -1, DNS: 8600)
Cluster Addr: 192.168.231.147 (LAN: 8301, WAN: 8302)
Encrypt: Gossip: false, TLS-Outgoing: false, TLS-Incoming: false
==> Log data will now stream in as it occurs:
2017/12/06 23:36:46 [INFO] serf: EventMemberJoin: n3 192.168.231.147
2017/12/06 23:36:46 [INFO] agent: Started DNS server 127.0.0.1:8600 (udp)
2017/12/06 23:36:46 [INFO] agent: Started DNS server 127.0.0.1:8600 (tcp)
2017/12/06 23:36:46 [INFO] agent: Started HTTP server on 127.0.0.1:8500 (tcp)
2017/12/06 23:36:46 [INFO] agent: started state syncer
2017/12/06 23:36:46 [WARN] manager: No servers available
2017/12/06 23:36:46 [ERR] agent: failed to sync remote state: No known Consul servers
2017/12/06 23:37:08 [WARN] manager: No servers available
2017/12/06 23:37:08 [ERR] agent: failed to sync remote state: No known Consul servers
2017/12/06 23:37:36 [WARN] manager: No servers available
2017/12/06 23:37:36 [ERR] agent: failed to sync remote state: No known Consul servers
2017/12/06 23:38:02 [WARN] manager: No servers available
2017/12/06 23:38:02 [ERR] agent: failed to sync remote state: No known Consul servers
2017/12/06 23:38:22 [WARN] manager: No servers available
2017/12/06 23:38:22 [ERR] agent: failed to sync remote state: No known Consul servers
在n3上面将节点n3加入集群
[join 192.168.231.145 ]$ ./consul
再次查看集群节点信息
此时三个节点的Consul集群搭建成功了!其实n1和n2是Server模式启动,n3是Client模式启动。
关于Consul的Server模式和Client模式主要的区别是这样的,一个Consul集群通过启动的参数
-bootstrap-expect
来控制这个集群的Server节点个数,Server模式的节点会维护集群的状态,并且如果某个Server节点退出了集群,则会触发Leader重新选举机制,在会剩余的Server模式节点中重新选举一个Leader;而Client模式的节点的加入和退出很自由。在n2中启动web-ui
2. 一个空格引发的“惨案“
3. 大型网站架构演化发展历程
扫码二维码关注我
·end·
—如果本文有帮助,请分享到朋友圈吧—
我们一起愉快的玩耍!