etcd 安装与管理

比眉伴天荒 2022-06-03 05:50 402阅读 0赞

# 目的 #

初始化启动 etcd 服务
    检测 etcd 集群健康状态
    检测 etcd 集群成员
    重置 etcd 集群

# 环境说明 #

<table> 
 <tbody> 
  <tr> 
   <td>角色</td> 
   <td>主机名</td> 
   <td>ipaddr</td> 
   <td>os</td> 
   <td>etcd 版本</td> 
  </tr> 
  <tr> 
   <td>etcd</td> 
   <td>qemu-test6.vclound.com</td> 
   <td>xx.xxx.205.229</td> 
   <td>centos 7</td> 
   <td>etcd-3.0.15-1.el7.x86_64</td> 
  </tr> 
  <tr> 
   <td>etcd</td> 
   <td>qemu-test3.vclound.com</td> 
   <td>xx.xxx.205.226</td> 
  </tr> 
  <tr> 
   <td>etcd</td> 
   <td>qemu-test8.vclound.com</td> 
   <td>xx.xxx.205.231</td> 
  </tr> 
 </tbody> 
</table>

# 软件安装 #

yum install -y etcd

# 配置 #

每个 member 根据自己主机名, IP 地址, 配置相应  
/etc/etcd/etcd.conf

[root@qemu-test6 ~]# grep -v "#" /etc/etcd/etcd.conf
    ETCD_NAME=qemu-test6
    ETCD_DATA_DIR="/var/lib/etcd/default.etcd"
    ETCD_LISTEN_PEER_URLS="http://xx.xxx.205.229:2380"
    ETCD_LISTEN_CLIENT_URLS="http://xx.xxx.205.229:2379,http://127.0.0.1:2379"
    ETCD_INITIAL_ADVERTISE_PEER_URLS="http://xx.xxx.205.229:2380"
    ETCD_INITIAL_CLUSTER="qemu-test6=http://xx.xxx.205.229:2380,qemu-test3=http://xx.xxx.205.226:2380,qemu-test8=http://xx.xxx.205.231:2380"
    ETCD_INITIAL_CLUSTER_STATE="new"
    ETCD_INITIAL_CLUSTER_TOKEN="etcd-skydns-cluster"
    ETCD_ADVERTISE_CLIENT_URLS="http://xx.xxx.205.229:2379"

## 详细说明 ##

### 主机名配置 ###

ETCD_NAME
    每个机器填写自己主机名, 取 hostname -s 值即可,

### 监听地址 ###

ETCD_LISTEN_CLIENT_URLS="http://xx.xxx.205.229:2379,http://127.0.0.1:2379"
    用于 etcdctl 命令连接, 其中 127.0.01 用于本地连接
    
    ETCD_LISTEN_PEER_URLS="http://xx.xxx.205.229:2380"
    用于监听其他 etcd member 连接

### 初始化 ###

ETCD_INITIAL_ADVERTISE_PEER_URLS="http://xx.xxx.205.229:2380"
    本地用于监听并连接其他 member 的地址
    
    ETCD_INITIAL_CLUSTER="qemu-test6=http://xx.xxx.205.229:2380,qemu-test3=http://xx.xxx.205.226:2380,qemu-test8=http://xx.xxx.205.231:2380"
    启动集群使, 使用静态连接方法, 定义每个 member 主机名 endpoint
    
    ETCD_INITIAL_CLUSTER_STATE="new"
    用于表示初始化集群
    
    ETCD_INITIAL_CLUSTER_TOKEN="etcd-skydns-cluster"  
    用于标记集群唯一性的 token

### 启动 etcd ###

注意, 第一台启动的 etcd 服务器将会不断等待其他集群成员完成初始化连接, 处于 waiting 状态

systemctl start etcd

### 服务检测状态 ###

[root@qemu-test6 ~]# systemctl status etcd
    ● etcd.service - Etcd Server
       Loaded: loaded (/usr/lib/systemd/system/etcd.service; disabled; vendor preset: disabled)
       Active: active (running) since Mon 2017-12-11 10:57:52 CST; 11s ago
     Main PID: 10135 (etcd)
       CGroup: /system.slice/etcd.service
               └─10135 /usr/bin/etcd --name=qemu-test6 --data-dir=/var/lib/etcd/default.etcd --listen-client-urls=http://xx.xxx.205.229:2379
    
    Dec 11 10:57:56 qemu-test6 etcd[10135]: failed to reach the peerURL(http://xx.xxx.205.231:2380) of member f8ef84dbaeb0e89a (Get http://xx.xxx.205.231:2380/vers...on refused)
    Dec 11 10:57:56 qemu-test6 etcd[10135]: cannot get the version of member f8ef84dbaeb0e89a (Get http://xx.xxx.205.231:2380/version: dial tcp xx.xxx.205.231:2380...on refused)
    Dec 11 10:57:59 qemu-test6 etcd[10135]: peer f8ef84dbaeb0e89a became active
    Dec 11 10:57:59 qemu-test6 etcd[10135]: established a TCP streaming connection with peer f8ef84dbaeb0e89a (stream MsgApp v2 writer)
    Dec 11 10:57:59 qemu-test6 etcd[10135]: established a TCP streaming connection with peer f8ef84dbaeb0e89a (stream Message writer)
    Dec 11 10:57:59 qemu-test6 etcd[10135]: established a TCP streaming connection with peer f8ef84dbaeb0e89a (stream Message reader)
    Dec 11 10:57:59 qemu-test6 etcd[10135]: established a TCP streaming connection with peer f8ef84dbaeb0e89a (stream MsgApp v2 reader)
    Dec 11 10:58:00 qemu-test6 etcd[10135]: updating the cluster version from 2.3 to 3.0
    Dec 11 10:58:00 qemu-test6 etcd[10135]: updated the cluster version from 2.3 to 3.0
    Dec 11 10:58:00 qemu-test6 etcd[10135]: enabled capabilities for version 3.0
    Hint: Some lines were ellipsized, use -l to show in full.

### 关闭 etcd ###

systemctl stop etcd

### 集群健康检测 ###

[root@qemu-test6 ~]# etcdctl cluster-health
    member 3a412437d7a67c64 is healthy: got healthy result from http://xx.xxx.205.229:2379
    member eb6747efc5a87825 is healthy: got healthy result from http://xx.xxx.205.226:2379
    member f8ef84dbaeb0e89a is healthy: got healthy result from http://xx.xxx.205.231:2379
    cluster is healthy

### 集群成员检测 ###

[root@qemu-test6 ~]# etcdctl member list
    3a412437d7a67c64: name=qemu-test6 peerURLs=http://xx.xxx.205.229:2380 clientURLs=http://xx.xxx.205.229:2379 isLeader=true
    eb6747efc5a87825: name=qemu-test3 peerURLs=http://xx.xxx.205.226:2380 clientURLs=http://xx.xxx.205.226:2379 isLeader=false
    f8ef84dbaeb0e89a: name=qemu-test8 peerURLs=http://xx.xxx.205.231:2380 clientURLs=http://xx.xxx.205.231:2379 isLeader=false

#### 常见错误 ####

[root@qemu-test6 ~]# etcdctl member list
    Error:  client: etcd cluster is unavailable or misconfigured
    error #0: dial tcp 127.0.0.1:2379: getsockopt: connection refused
    error #1: dial tcp 127.0.0.1:4001: getsockopt: connection refused

上述问题, 是 ETCD\_LISTEN\_CLIENT\_URLS=”http://127.0.0.1:2379” 没有提供 127 本地地址导致,那么通过指定 endpoint 方法解决

etcdctl --endpoints http://xx.xxx.205.229:2379,http://xx.xxx.205.229:4001 member list

## 集群初重置 ##

1 利用 systemctl stop etcd 关闭集群, 在重新启动集群, 之前 etcd 保存的资料不会丢失
    2 要对 etcd 集群进行重置, 最简单的方式是关闭集群后, 
    3 删除了所有 etcd member 中 ETCD_DATA_DIR 配置中定义的目录下的所有子目录
    4 启动所有 member 即可