本文摘自：http://blog.csdn.net/tzm1921995/article/details/52212168
对于通过apt或yum方式安装etcd的，可以修改etcd.conf配置文件来配置各个参数。
Etcd集群简介
etcd组件作为一个高可用强一致性的服务发现存储仓库。我们此次在三个主机上搭建了一个包含三个Etcd节点的集群，实现了集群的动态扩展和收缩，并测试和验证了Etcd集群键——值存储的一致性和高
一、环境准备
OS: Ubuntu 14.04
user: root
ipaddress:
etcd01: 192.168.200.24
etcd02: 192.168.200.25
etcd03: 192.168.200.26
下载etcd 源码包：https://github.com/coreos/etcd/releases/
这里用的是etcd-v2.3.7-Linux-amd64.tar.gz
二、安装配置etcd
etcd01上安装
tar xf etcd-v2.3.7-linux-amd64.tar.gz
cd etcd-v2.3.7-linux-amd64
cp etcd* /usr/local/bin/
创建etcd01脚本，可方便配置etcd启动参数
#cat etcd01
/usr/local/bin/etcd -name etcd01 \
-data-dir /data/etcd01 \
-advertise-client-urls http://192.168.200.24:2379,http://192.168.200.24:4001 \
-listen-client-urls http://0.0.0.0:2379,http://192.168.200.24:4001 \
-initial-advertise-peer-urls http://192.168.200.24:2380 \
-listen-peer-urls http://0.0.0.0:2380 \
-initial-cluster-token etcd-cluster-1 \
-initial-cluster etcd01=http://192.168.200.24:2380,etcd02=http://192.168.200.25:2380,etcd03=http://192.168.200.26:2380 \
-initial-cluster-state new
参数说明：-name 指定名字
-data-dir 指定数据保存目录，默认是当前目录
-initial-cluster-state 集群状态 new为新创建集群 existing为已存在（可不指定）
在etcd02 etcd03上分别做相似操作
脚本中-advertise-client-urls 和 -initial-advertis-peer-urls 参数修改一下即可
然后分别运行脚本：nohup ./etcd01 &
三、测试
在任一台主机上执行etcdctl member list
#etcdctl member list
6a223770249e927d: name=etcd02 peerURLs=http://192.168.200.25:2380 clientURLs=http://192.168.200.25:2379,http://192.168.200.25:4001 isLeader=false
7e0ce16121dfea24: name=etcd01 peerURLs=http://192.168.200.24:2380 clientURLs=http://192.168.200.24:2379,http://192.168.200.24:4001 isLeader=true
bfc28be8765b503e: name=etcd03 peerURLs=http://192.168.200.26:2380 clientURLs=http://192.168.200.26:2379,http://192.168.200.26:4001 isLeader=false
可以看到集群的节点情况，并能看出哪个是leader节点
我们在etcd01上设置一个key/value
root@etcd1:~# etcdctl set api_server http://192.168.5.44:8080
http://192.168.5.44:8080
这时就可以在任意一台主机上获取这个key/value
root@etcd2:~# etcdctl get api_server
root@etcd3:~# etcdctl get api_server
http://192.168.5.44:8080
在member list上看到etcd01是leader ，这时把etcd01停掉（kill）
用etcdctl cluster-health查看
root@etcd2:~# etcdctl cluster-health
member 6a223770249e927d is healthy: got healthy result from http://192.168.200.25:2379
failed to check the health of member 7e0ce16121dfea24 on http://192.168.200.24:2379: Get http://192.168.200.24:2379/health: dial tcp 192.168.200.24:2379: getsockopt: connection refused
failed to check the health of member 7e0ce16121dfea24 on http://192.168.200.24:4001: Get http://192.168.200.24:4001/health: dial tcp 192.168.200.24:4001: getsockopt: connection refused
member 7e0ce16121dfea24 is unreachable: [http://192.168.200.24:2379 http://192.168.200.24:4001] are all unreachable
member bfc28be8765b503e is healthy: got healthy result from http://192.168.200.26:2379
cluster is healthy
并且集群leader进行了重新选举
root@etcd2:~# etcdctl member list
6a223770249e927d: name=etcd02 peerURLs=http://192.168.200.25:2380 clientURLs=http://192.168.200.25:2379,http://192.168.200.25:4001 isLeader=true
7e0ce16121dfea24: name=etcd01 peerURLs=http://192.168.200.24:2380 clientURLs=http://192.168.200.24:2379,http://192.168.200.24:4001 isLeader=false
bfc28be8765b503e: name=etcd03 peerURLs=http://192.168.200.26:2380 clientURLs=http://192.168.200.26:2379,http://192.168.200.26:4001 isLeader=false
现在etcd02是leader了，这时我们在群集中设置两个key/value
root@etcd2:~# etcdctl set test01 123456
123456
root@etcd2:~# etcdctl set test02 abcdefg
abcdefg
重新启动etcd01
root@etcd1:~# etcdctl cluster-health
member 6a223770249e927d is healthy: got healthy result from http://192.168.200.25:2379
member 7e0ce16121dfea24 is healthy: got healthy result from http://192.168.200.24:2379
member bfc28be8765b503e is healthy: got healthy result from http://192.168.200.26:2379
cluster is healthy
root@etcd1:~# etcdctl get test01
123456
root@etcd1:~# etcdctl get test02
abcdefg
但这时在etcd01重新加入集群，并保持了key/value的全局一致性，由此可见 etcd 搭建的集群是可以实现高可用的。
Etcd集群的扩展与收缩
etcd集群如果收缩很简单，直接在命令行输入
etcdctl member remove {$memberID}
$memberID是你即将要删除节点的etcd的ID，etcd的扩展有一些地方需要注意一下，我在这里操作的时候遇到了不少坑。从上文写到现在，有一个文件夹很重要，几乎每个坑都与它有关，那就是-data-dir所声明的文件夹，注意要扩展一个etcd集群时，首先在集群内的任一台机器上输入
etcdctl member add $etcd_name $peer_url
$etcd_name：新加入的etcd节点的名字
$peer_url：一般为新加入的节点 IP:2380
下面为通过yum方式安装后，/etc/etcd/etcd.conf配置文件的参考实例：
# 节点名称
ETCD_NAME=etcd0
# # 数据存放位置
ETCD_DATA_DIR="/var/lib/etcd/default.etcd"
# # 监听其他 Etcd 实例的地址
ETCD_LISTEN_PEER_URLS="http://0.0.0.0:2380"
# # 监听客户端地址
ETCD_LISTEN_CLIENT_URLS="http://0.0.0.0:2379,http://0.0.0.0:4001"
# # 通知其他 Etcd 实例地址
ETCD_INITIAL_ADVERTISE_PEER_URLS="http://192.168.212.140:2380"
# # 初始化集群内节点地址
ETCD_INITIAL_CLUSTER="etcd0=http://192.168.212.140:2380,etcd1=http://192.168.212.139:2380,etcd2=http://192.168.212.138:2380"
# # 初始化集群状态，new 表示新建
ETCD_INITIAL_CLUSTER_STATE="new"
# # 初始化集群 token
ETCD_INITIAL_CLUSTER_TOKEN="mritd-etcd-cluster"
# # 通知 客户端地址
ETCD_ADVERTISE_CLIENT_URLS="http://192.168.212.140:2379,http://192.168.212.140:4001"
