上一篇文章，顺利完成了在CentOS7上安装部署kubernetes。接下来，考虑在CentOS7配置部署etcd的高可用集群。

从etcd clustering的官方站点上，看到有3中方式来引导etcd cluster。这里分别以静态引导和public etcd discovery这两种方式给出配置过程。假定：

• 你已经在CentOS7上安装好etcd，如果没有，可以参考在CentOS 7上安装部署kubernetes，或者利用yum单独安装etcd;
• 要配置etcd cluster的3台机器分别为：172.16.11.30，172.16.11.71，172.16.11.78；
• 及其在cluster中的命名依次为：infra0,infra1,infra2;

一 采用static方式开始执行配置：

1 第1台机器172.16.11.30上执行：

[root@localhost ~]# etcd --name infra0 --initial-advertise-peer-urls http://172.16.11.30:2380 --listen-peer-urls http://172.16.11.30:2380 --listen-client-urls http://172.16.11.30:2379,http://127.0.0.1:2379 --advertise-client-urls http://172.16.11.30:2379 --initial-cluster-token etcd-cluster-1 --initial-cluster infra0=http://172.16.11.30:2380,infra1=http://172.16.11.71:2380,infra2=http://172.16.11.78:2380 --initial-cluster-state new

2 第2台机器172.16.11.71上执行：

[root@test-zyd-jy-2 ~]# etcd --name infra1 --initial-advertise-peer-urls http://172.16.11.71:2380 --listen-peer-urls http://172.16.11.71:2380 --listen-client-urls http://172.16.11.71:2379,http://127.0.0.1:2379 --advertise-client-urls http://172.16.11.71:2379 --initial-cluster-token etcd-cluster-1 --initial-cluster infra0=http://172.16.11.30:2380,infra1=http://172.16.11.71:2380,infra2=http://172.16.11.78:2380 --initial-cluster-state new

3 第3台机器172.16.11.78上执行：

[root@localhost ~]# etcd --name infra2 --initial-advertise-peer-urls http://172.16.11.78:2380 --listen-peer-urls http://172.16.11.78:2380 --listen-client-urls http://172.16.11.78:2379,http://127.0.0.1:2379 --advertise-client-urls http://172.16.11.78:2379 --initial-cluster-token etcd-cluster-1 --initial-cluster infra0=http://172.16.11.30:2380,infra1=http://172.16.11.71:2380,infra2=http://172.16.11.78:2380 --initial-cluster-state new

4 任意机器上，验证cluster健康信息：

[root@localhost ~]# etcdctl cluster-health
member 2a21240027dcd247 is healthy: got healthy result from http://172.16.11.30:2379
member 784c02b20039dc5b is healthy: got healthy result from http://172.16.11.71:2379
member b1ca771041e5f776 is healthy: got healthy result from http://172.16.11.78:2379
cluster is healthy
[root@localhost ~]#

5 任意机器上，执行etcdctl set key value：

[root@localhost ~]# etcdctl set k1 100
100
[root@localhost ~]# etcdctl get k1
100
[root@localhost ~]#

6 其它机器上，执行etcdctl get key来验证：

[root@test-zyd-jy-2 ~]# etcdctl get k1
100
[root@test-zyd-jy-2 ~]# 

至此，通过静态方式，配置了一个3个节点的etcd cluster。关于上述命令中，几个参数的说明：

• name:指位于同一个cluster中的每个etcd 节点的名字，必须唯一；
• listen-client-urls：etcd接收客户端发起请求的监听地址；
• advertise-client-urls：etcd cluster中的成员用于给其它成员或者客户端，发送广播的地址信息；该地址既不能配置为localhost的形式，也绝对不能留空；

 

二 采用discovery方式来配置etcd cluster

discovery方式配置etcd cluster又分为etcd discovery service和DNS discovery方式，这里采用etcd discovery service方式来配置etcd cluster。

1 创建public discovery的token

curl https://discovery.etcd.io/new?size=3

这里，只需在其中任意节点执行上述命令，并获取命令的结果。这个结果集用于整个etcd cluster的配置。

 [root@localhost ~]# curl https://discovery.etcd.io/new?size=3
https://discovery.etcd.io/920771e9b700844194c13ebeff79a543[root@localhost ~]# 

2 第1台机器172.16.11.30上执行：

[root@localhost ~]# etcd --name centos-master --initial-advertise-peer-urls http://172.16.11.30:2380 --listen-peer-urls http://172.16.11.30:2380  --listen-client-urls http://127.0.0.1:4001,http://172.16.11.30:4001 --advertise-client-urls http://172.16.11.30:4001  --discovery https://discovery.etcd.io/920771e9b700844194c13ebeff79a543
2018-01-16 09:40:51.420704 I | etcdmain: etcd Version: 3.2.9
2018-01-16 09:40:51.421594 I | etcdmain: Git SHA: f1d7dd8
2018-01-16 09:40:51.421624 I | etcdmain: Go Version: go1.8.3
2018-01-16 09:40:51.421644 I | etcdmain: Go OS/Arch: linux/amd64
2018-01-16 09:40:51.421673 I | etcdmain: setting maximum number of CPUs to 4, total number of available CPUs is 4
2018-01-16 09:40:51.422265 W | etcdmain: no data-dir provided, using default data-dir ./centos-master.etcd
2018-01-16 09:40:51.422854 I | embed: listening for peers on http://172.16.11.30:2380
2018-01-16 09:40:51.423539 I | embed: listening for client requests on 127.0.0.1:4001
2018-01-16 09:40:51.423694 I | embed: listening for client requests on 172.16.11.30:4001
2018-01-16 09:40:53.344420 N | discovery: found self 425cf1a6e9b4c574 in the cluster
2018-01-16 09:40:53.344515 N | discovery: found 1 peer(s), waiting for 2 more
...
...

3  第1台机器172.16.11.71上执行：

[root@test-zyd-jy-2 ~]# etcd --name centos-master71 --initial-advertise-peer-urls http://172.16.11.71:2380 --listen-peer-urls http://172.16.11.71:2380  --listen-client-urls http://127.0.0.1:4001,http://172.16.11.71:4001 --advertise-client-urls http://172.16.11.71:4001  --discovery https://discovery.etcd.io/920771e9b700844194c13ebeff79a543
2018-01-16 09:42:18.030679 I | etcdmain: etcd Version: 3.2.9
2018-01-16 09:42:18.030858 I | etcdmain: Git SHA: f1d7dd8
2018-01-16 09:42:18.030872 I | etcdmain: Go Version: go1.8.3
2018-01-16 09:42:18.030884 I | etcdmain: Go OS/Arch: linux/amd64
2018-01-16 09:42:18.030897 I | etcdmain: setting maximum number of CPUs to 4, total number of available CPUs is 4
2018-01-16 09:42:18.030920 W | etcdmain: no data-dir provided, using default data-dir ./centos-master71.etcd
2018-01-16 09:42:18.031763 N | etcdmain: the server is already initialized as member before, starting as etcd member...
2018-01-16 09:42:18.032013 I | embed: listening for peers on http://172.16.11.71:2380
2018-01-16 09:42:18.032364 I | embed: listening for client requests on 127.0.0.1:4001
2018-01-16 09:42:18.032463 I | embed: listening for client requests on 172.16.11.71:4001
2018-01-16 09:42:20.062578 N | discovery: found peer 425cf1a6e9b4c574 in the cluster
2018-01-16 09:42:20.062701 N | discovery: found self d91473943b320695 in the cluster
2018-01-16 09:42:20.062718 N | discovery: found 2 peer(s), waiting for 1 more

4  第1台机器172.16.11.78上执行：

[root@localhost ~]# etcd --name centos-master78 --initial-advertise-peer-urls http://172.16.11.78:2380 --listen-peer-urls http://172.16.11.78:2380  --listen-client-urls http://127.0.0.1:4001,http://172.16.11.78:4001 --advertise-client-urls http://172.16.11.78:4001  --discovery https://discovery.etcd.io/920771e9b700844194c13ebeff79a543
2018-01-16 09:43:08.553406 I | etcdmain: etcd Version: 3.2.9
2018-01-16 09:43:08.553770 I | etcdmain: Git SHA: f1d7dd8
2018-01-16 09:43:08.553785 I | etcdmain: Go Version: go1.8.3
2018-01-16 09:43:08.553799 I | etcdmain: Go OS/Arch: linux/amd64
2018-01-16 09:43:08.553812 I | etcdmain: setting maximum number of CPUs to 8, total number of available CPUs is 8
2018-01-16 09:43:08.553840 W | etcdmain: no data-dir provided, using default data-dir ./centos-master78.etcd
2018-01-16 09:43:08.554396 N | etcdmain: the server is already initialized as member before, starting as etcd member...
2018-01-16 09:43:08.554762 I | embed: listening for peers on http://172.16.11.78:2380
2018-01-16 09:43:08.554951 I | embed: listening for client requests on 127.0.0.1:4001
2018-01-16 09:43:08.555013 I | embed: listening for client requests on 172.16.11.78:4001
2018-01-16 09:43:10.345677 N | discovery: found peer 425cf1a6e9b4c574 in the cluster
2018-01-16 09:43:10.345813 N | discovery: found peer d91473943b320695 in the cluster
2018-01-16 09:43:10.345831 N | discovery: found self 98d67d53103d0e5e in the cluster
2018-01-16 09:43:10.345845 N | discovery: found 3 needed peer(s)
2018-01-16 09:43:10.346126 I | etcdserver: name = centos-master78
2018-01-16 09:43:10.346145 I | etcdserver: data dir = centos-master78.etcd
2018-01-16 09:43:10.346161 I | etcdserver: member dir = centos-master78.etcd/member
2018-01-16 09:43:10.346210 I | etcdserver: heartbeat = 100ms
2018-01-16 09:43:10.346226 I | etcdserver: election = 1000ms
2018-01-16 09:43:10.346240 I | etcdserver: snapshot count = 100000
2018-01-16 09:43:10.346286 I | etcdserver: discovery URL= https://discovery.etcd.io/920771e9b700844194c13ebeff79a543
2018-01-16 09:43:10.346322 I | etcdserver: advertise client URLs = http://172.16.11.78:4001
2018-01-16 09:43:10.346368 I | etcdserver: initial advertise peer URLs = http://172.16.11.78:2380
2018-01-16 09:43:10.346429 I | etcdserver: initial cluster = centos-master78=http://172.16.11.78:2380
2018-01-16 09:43:10.353143 I | etcdserver: starting member 98d67d53103d0e5e in cluster 57d019a9b13bd63a
2018-01-16 09:43:10.353268 I | raft: 98d67d53103d0e5e became follower at term 0

5  任意机器上，验证cluster健康信息：

[root@localhost ~]# etcdctl cluster-health
member 425cf1a6e9b4c574 is healthy: got healthy result from http://172.16.11.30:4001
member 98d67d53103d0e5e is healthy: got healthy result from http://172.16.11.78:4001
member d91473943b320695 is healthy: got healthy result from http://172.16.11.71:4001
cluster is healthy
[root@localhost ~]# 

至此，通过static和discovery service两种方式来配置了etcd cluster。