@H_403_3@
1.配置@H_403_3@ RHCS 集群的前提：@H_403_3@
时间同步@H_403_3@
@H_403_3@
2. RHCS 所需要的主要软件包为@H_403_3@cman和@H_403_3@ rgmanager
cman: 是集群基础信息层，在@H_403_3@ CentOS 6中依赖@H_403_3@ corosync
rgmanager: 是集群资源管理器，@H_403_3@ 类似于@H_403_3@pacemaker 的功能@H_403_3@
luci: 提供了管理@H_403_3@ rhcs 集群的@H_403_3@ web 界面@H_403_3@,luci 管理集群主要是通过跟@H_403_3@ ricci 通信来完成的。@H_403_3@
ricci: 安装在集群的节点的接收来自@H_403_3@ luci 管理请求的代理。@H_403_3@
luci 跟@H_403_3@ ricci 的关系就好像@H_403_3@ ambari-server 跟@H_403_3@ ambari-agent 一样。@H_403_3@
3.环境说明@H_403_3@:
luci:192.168.6.31cent1.test.com ricci:192.168.6.32cent2.test.com ricci:192.168.6.33cent3.test.com ricci:192.168.6.34cent4.test.com
我这里已经配好了主机名了，但是其他的如时间同步，配置@H_403_3@/etc/hosts/ 等都没执行，为了方便，所以写了个@H_403_3@ playbook 来进行初始化一下@H_403_3@
---
-hosts:hdpservers
remote_user:root
vars:
tasks:
-name:addsynctimecron
cron:name='synctime'minute='*/5'job='/usr/sbin/ntpdate192.168.6.31'
-name:shutdowniptables
service:name={{item.name}}state={{item.state}}enabled={{item.enabled}}
with_items:
-{name:iptables,state:stopped,enabled:no}
-{name:NetworkManager,enabled:no}
tags:stopservice
-name:copyselinuxconffile
copy:src={{item.src}}dest={{item.dest}}owner={{item.owner}}group={{item.group}}mode={{item.mode}}
with_items:
-{src=\'#\'" /etc/selinux/config',dest:/etc/selinux/config,owner:root,group:root,mode:'0644'}
-{src=\'#\'" /etc/hosts',dest:/etc/hosts,mode:'0644'}
-name:cmdoffselinux
shell:setenforce0
执行这个 playbook，进行初始化
[root@cent1yaml]#ansible-playbookbase.yml
4.在@H_403_3@ cent1 上安装@H_403_3@ luci,luci 是一个@H_403_3@ python 程序，依赖很多@H_403_3@python包@H_403_3@
[root@cent1~]#yuminstallluci
[root@cent3~]#/etc/init.d/lucistart Addingfollowingauto-detectedhostIDs(IPaddresses/domainnames),correspondingto`cent3'address,totheconfigurationofself-managedcertificate`/var/lib/luci/etc/cacert.config'(youcanchangethembyediting`/var/lib/luci/etc/cacert.config',removingthegeneratedcertificate`/var/lib/luci/certs/host.pem'andrestartingluci): (nonesuitablefound,youcanstilldoitmanuallyasmentionedabove) Generatinga2048bitRSAprivatekey writingnewprivatekeyto'/var/lib/luci/certs/host.pem' 正在启动saslauthd：[确定] Startluci...[确定] Pointyourwebbrowsertohttps://cent1.hfln.com:8084(orequivalent)toaccessluci
现在可以在前台登录@H_403_3@luci 了，看清是@H_403_3@ https 哦@H_403_3@
账号密码就是这台主机的账号和密码@H_403_3@
登录成功啦@H_403_3@,@H_403_3@现在来配置@H_403_3@ rhcs @H_403_3@的集群，这个只是用来管理集群的，@H_403_3@真正的集群还没开始装呢。@H_403_3@
5.在@H_403_3@ cnet2,cent3,cent4 @H_403_3@中安装@H_403_3@ ricci,ricci @H_403_3@也依赖很多软件，这里使用@H_403_3@ ansible @H_403_3@直接在三个节点上装，@H_403_3@ @H_403_3@当然我已经配好了@H_403_3@ cent1 @H_403_3@到@H_403_3@ @H_403_3@其他节点的免密钥登录了@H_403_3@@H_403_3@
[root@cent1~]#ansiblerhcs-myum-a"name=ricci"
装好@H_403_3@ricci @H_403_3@之后还要在@H_403_3@ node @H_403_3@节点上给@H_403_3@ ricci @H_403_3@用户设置密码，@H_403_3@ricci@H_403_3@用户就是运行@H_403_3@ ricci@H_403_3@进程的用户，这个密码一会要用@H_403_3@,@H_403_3@这里就简单粗暴了，这个密码还可以用@H_403_3@ ccs@H_403_3@命令来进行设置@H_403_3@
[root@cent1~]#ansiblerhcs-mshell-a"echo'123456'|passwd--stdinricci"
@H_403_3@
[root@cent1~]#ansiblerhcs-mservice-a"name=riccistate=startedenabled=yes"
[root@cent2~]#ss-tunlp|grepricci
tcpLISTEN05:::11111:::*users:(("ricci",3237,3))
@H_403_3@
ricci @H_403_3@监听在@H_403_3@ 11111 @H_403_3@端口,像这种操作当然也是可以写到@H_403_3@ playbook @H_403_3@当中的@H_403_3@
6. @H_403_3@现在可以在@H_403_3@web @H_403_3@界面上配置@H_403_3@集群了，比如创建@H_403_3@/@H_403_3@添加@H_403_3@/@H_403_3@删除一个集群，管理@H_403_3@node,resource,fence device,servicegroups,Failover Domains @H_403_3@等等集群的全生命周期都可以在这里完成。@H_403_3@
@H_403_3@
这里演示一个关于@H_403_3@ web@H_403_3@服务的高可用服务@H_403_3@
Manage Clusters--> Create@H_403_3@ @H_403_3@是创建一个集群@H_403_3@
这个界面还算简单吧;@H_403_3@
Create Cluster @H_403_3@之后，那么就开始尝试安装集群软件了@H_403_3@.@H_403_3@
在任意一个@H_403_3@node@H_403_3@上可以看到@H_403_3@ ricci @H_403_3@的工作进程:@H_403_3@
[root@cent2~]#psaux|grepricci ricci34530.10.42136644400?S<s17:180:00ricci-uricci ricci34890.00.1549121908?S<s17:220:00/usr/libexec/ricci/ricci-worker-f/var/lib/ricci/queue/1500004777 root34900.20.5485525136?S17:220:00ricci-modrpm root35670.00.0103252880pts/0S+17:240:00grepricci
@H_403_3@
/var/lib/ricci/queue/@H_403_3@目录下存放的是@H_403_3@ luci @H_403_3@发给@H_403_3@ ricci @H_403_3@的任务文件，是@H_403_3@ XML @H_403_3@格式的@H_403_3@
[root@cent2~]#file/var/lib/ricci/queue/1500004777 /var/lib/ricci/queue/1500004777:XMLdocumenttext
可以点任何一个@H_403_3@node @H_403_3@进去看看@H_403_3@
如果这底下的服务没启动的话，可以尝试手动起一下，一般来说是@H_403_3@OK@H_403_3@的。@H_403_3@
这里没有@H_403_3@ fence @H_403_3@设备，不关注这个，添加两个公共资源，并添加一个服务，然后来启动服务@H_403_3@
Resources -->Add @H_403_3@: @H_403_3@添加一个资源@H_403_3@
添加一个虚拟@H_403_3@IP@H_403_3@，这里的@H_403_3@ mask @H_403_3@要写成上面这样，不能写成@H_403_3@ 255.255.255.0 @H_403_3@这种，否则会导致无法添加@H_403_3@IP@H_403_3@
rgmanagerStartingstoppedserviceservice:web1 rgmanagerstartonip"192.168.6.100/255.255.255.0"returned1(genericerror) rgmanager#68:Failedtostartservice:web1;returnvalue:1
再添加一个@H_403_3@script@H_403_3@资源@H_403_3@
9.@H_403_3@添加@H_403_3@ Service@H_403_3@
这里的资源是共公的，假如这个集群内有多个服务，那么都可以使用这些资源，也可以在@H_403_3@
Service Groups @H_403_3@里添加一个私有的资源。@H_403_3@
现在添加一个@H_403_3@Service:@H_403_3@
@H_403_3@
Service Groups--> Add @H_403_3@: @H_403_3@添加一个@H_403_3@ Service,@H_403_3@
@H_403_3@
Add Resource @H_403_3@将刚才建立的两个资源添加进来;@H_403_3@@H_403_3@
现在在@H_403_3@集群的节点上用命令查看一下，集群内的任何节点都可以@H_403_3@
[root@cent3~]#clustat ClusterStatusforha1@SunJan817:47:402017 MemberStatus:Quorate MemberNameIDStatus -------------------- cent2.test.com1Online,rgmanager cent3.test.com2Online,Local,rgmanager cent4.test.com3Online,rgmanager ServiceNameOwner(Last)State --------------------------- service:web1cent2.test.comstarted
在@H_403_3@ cent2 @H_403_3@上@H_403_3@ ip @H_403_3@和@H_403_3@httpd @H_403_3@服务都已经起来了@H_403_3@
[root@cent2~]#ipa 1:lo:<LOOPBACK,UP,LOWER_UP>mtu65536qdiscnoqueuestateUNKNOWN link/loopback00:00:00:00:00:00brd00:00:00:00:00:00 inet127.0.0.1/8scopehostlo inet6::1/128scopehost valid_lftforeverpreferred_lftforever 2:eth0:<BROADCAST,MULTICAST,LOWER_UP>mtu1500qdiscpfifo_faststateUPqlen1000 link/ether00:0c:29:91:b3:11brdff:ff:ff:ff:ff:ff inet192.168.6.32/24brd192.168.6.255scopeglobaleth0 inet192.168.6.100/24scopeglobalsecondaryeth0 inet6fe80::20c:29ff:fe91:b311/64scopelink valid_lftforeverpreferred_lftforever [root@cent2~]#netstat-tunlp|grep80 tcp00:::80:::*LISTEN34901/httpd
@H_403_3@
10.测试故障转移@H_403_3@:
关于@H_403_3@ rhcs 中@H_403_3@ service 的健康状态检测@H_403_3@,可以通过@H_403_3@ /var/log/cluster/rgmanager.log 日志来查看@H_403_3@
Jan0818:56:59rgmanager[ip]Checking192.168.6.100/24,Level10 Jan0818:56:59rgmanager[ip]192.168.6.100/24presentoneth0 Jan0818:56:59rgmanager[ip]Linkforeth0:Detected Jan0818:56:59rgmanager[ip]Linkdetectedoneth0 Jan0818:56:59rgmanager[ip]Localpingto192.168.6.100succeeded
这里可以看到他会尝试查看和@H_403_3@ ping 192.168.6.100 ，这是针对@H_403_3@ IP 资源的检测方式@H_403_3@
Jan0818:55:49rgmanager[script]Executing/etc/rc.d/init.d/httpdstatus
上面是@H_403_3@ script 资源的检测方式则是仅仅去用脚本来执行@H_403_3@ status 参数。@H_403_3@
在我尝试将@H_403_3@/etc/init.d/httpd/ stop 后，日志出现了如下:@H_403_3@
Jan0818:56:59rgmanager[script]Executing/etc/rc.d/init.d/httpdstatus Jan0818:56:59rgmanager[script]script:http1:statusof/etc/rc.d/init.d/httpdFailed(returned3) #这里发现检测失败了 Jan0818:56:59rgmanagerstatusonscript"http1"returned1(genericerror) Jan0818:56:59rgmanagerStoppingserviceservice:web1 Jan0818:56:59rgmanager[script]Executing/etc/rc.d/init.d/httpdstop Jan0818:56:59rgmanager[ip]RemovingIPv4address192.168.6.100/24frometh0 #以上几步在这个节点停止了web1服务 Jan0818:57:09rgmanagerServiceservice:web1isrecovering Jan0818:57:14rgmanagerServiceservice:web1isnowrunningonmember2 #将web1服务在member2上恢复了，member2也就是cent3.test.com
@H_403_3@
查看转移后的集群状态：@H_403_3@
[root@cent3~]#clustat ClusterStatusforha1@SunJan820:25:262017 MemberStatus:Quorate MemberNameIDStatus -------------------- cent2.test.com1Online,rgmanager ServiceNameOwner(Last)State --------------------------- service:web1cent3.test.comstarted
@H_403_3@
@H_403_3@
如果这种@H_403_3@ script 的资源不符合你的需求，那么可以尝试@H_403_3@ apache 资源。即使你认为这种@H_403_3@ script 的资源检查方式过于简单，也可以在脚本里添加功能来达到你的目的。@H_403_3@
11.尝试关闭节点，查看 Service 转移情况：
在关掉@H_403_3@ cent3 之后，@H_403_3@service 转移到了@H_403_3@ cent4上@H_403_3@
[root@cent2~]#clustat ClusterStatusforha1@SunJan820:35:422017 MemberStatus:Quorate MemberNameIDStatus -------------------- cent2.test.com1Online,rgmanager cent3.test.com2Offline cent4.test.com3Online,rgmanager ServiceNameOwner(Last)State --------------------------- service:web1cent4.test.comstarted
接着关掉了@H_403_3@ cent4,Service 又转移到了@H_403_3@ cent2
[root@cent2~]#clustat ClusterStatusforha1@SunJan820:36:272017 MemberStatus:Quorate MemberNameIDStatus -------------------- cent2.test.com1Online,rgmanager cent3.test.com2Offline cent4.test.com3Online ServiceNameOwner(Last)State --------------------------- service:web1cent2.test.comstarted
这里的 cent4.test.com 仍然显示 Online 是因为正在关机当中，尚未真正关闭。
[root@cent2~]# Messagefromsyslogd@cent2atJan820:36:42... rgmanager[5685]:#1:QuorumDissolved
Jan0820:35:01rgmanagerMember2shuttingdown Jan0820:36:18rgmanagerMember3shuttingdown Jan0820:36:18rgmanagerStartingstoppedserviceservice:web1 Jan0820:36:18rgmanager[ip]Linkforeth0:Detected Jan0820:36:19rgmanager[ip]AddingIPv4address192.168.6.100/24toeth0 Jan0820:36:19rgmanager[ip]Pingingaddr192.168.6.100fromdeveth0 Jan0820:36:21rgmanager[ip]SendinggratuitousARP:192.168.6.10000:0c:29:91:b3:11brdff:ff:ff:ff:ff:ff Jan0820:36:22rgmanager[script]Executing/etc/rc.d/init.d/httpdstart Jan0820:36:22rgmanagerServiceservice:web1started Jan0820:36:42rgmanager#1:QuorumDissolved Messagefromsyslogd@cent2atJan820:36:42... rgmanager[5685]:#1:QuorumDissolved Jan0820:36:42rgmanager[script]Executing/etc/rc.d/init.d/httpdstop Jan0820:36:42rgmanager[ip]RemovingIPv4address192.168.6.100/24frometh0
服务停止了，这是因为@H_403_3@ 法定票数不足的原因@H_403_3@
[root@cent2~]#clustat Servicestatesunavailable:Operationrequiresquorum ClusterStatusforha1@SunJan820:37:002017 MemberStatus:Inquorate MemberNameIDStatus -------------------- cent2.test.com1Online,Local cent3.test.com2Offline cent4.test.com3Offline
