集群LVS+GFS+ISCSI+TOMCAT

作者: hosyp  
LVS是中国人发起的项目,真是意外呀!大家可以看http://www.douzhe.com/linuxtips/1665.html
  我是从最初的HA(高可用性)开始的,别人的例子是用VMWARE,可以做试验但不能实际应用,我又
没有光纤卡的Share Storage,于是就选用ISCSI,成功后又发现ISCSI+EXT3不能用于LVS,倒最后发
现GFS可用,我最终成功配成可实际应用的LVS,前后断断续续花了四个月,走了很多弯路。我花了
三天时间写下这篇文章,希望对大家有用。
  这里要感谢linuxfans.org、linuxsir.com、chinaunix.com以及其它很多网站,很多资料都是从
他们的论坛上找到的。  
参考文档及下载点
a.http://www.gyrate.org/misc/gfs.txt
b.http://www.redhat.com/docs/manuals/enterprise/RHEL-3-Manual/cluster-suite/index.html
  http://www.redhat.com/docs/manuals/csgfs/admin-guide/index.html
c.ftp://ftp.redhat.com/pub/redhat/linux/updates/enterprise/3ES/en/RHGFS/SRPMS
d.http://distro.ibiblio.org/pub/linux/distributions/caoslinux/centos/3.1/contrib/i386/RPMS/
                 
                 LVS结构图:
              eth0=10.3.1.101
            eth0:1=10.3.1.254
                Load Balance
                   Router
              eth1=192.168.1.71  
            eth1:1=192.168.1.1
                  |         |
                  |         |
                Real1      Real2
       eth0=192.168.1.68  eth0=192.168.1.67    
            (eth0 gateway=192.168.1.1)
       eth1=192.168.0.1---eth1=192.168.0.2
                  (双机互联线)
                       |
                       |
                      GFS
                     ISCSI
                Share storage   
              eth0=192.168.1.124
             
1.Setup ISCSI Server
  Server: PIII 1.4,512M, Dell 1650,Redhat 9,IP=192.168.1.124
  从http://iscsitarget.sourceforge.net/下载ISCSI TARGET的Source code
  (http://sourceforge.net/project/showfiles.php?group_id=108475&package_id=117141)
  我选了iscsitarget-0.3.8.tar.gz,要求kernel 2.4.29
  从kernel.org下载kernel 2.4.29,解开编译重启后编译安装iscsitarget-0.3.8:
  #make KERNELSRC=/usr/src/linux-2.4.29
  #make KERNELSRC=/usr/src/linux-2.4.29 install
  #cp ietd.conf /etc
  #vi /etc/ietd.conf
    # Example iscsi target configuration
    #
    # Everything until the first target definition belongs
    # to the global configuration.
    # Right now this is only the user configuration used
    # during discovery sessions:
    # Users, who can access this target
    # (no users means anyone can access the target)
    User iscsiuser 1234567890abc
    Target iqn.2005-04.com.my:storage.disk2.sys1.iraw1
            User iscsiuser 1234567890abc
            Lun 0 /dev/sda5 fileio
            Alias iraw1
    Target iqn.2005-04.com.my:storage.disk2.sys1.iraw2
            User iscsiuser 1234567890abc
            Lun 1 /dev/sda6 fileio
            Alias iraw2
    Target iqn.2005-04.com.my:storage.disk2.sys2.idisk
            User iscsiuser 1234567890abc
            Lun 2 /dev/sda3 fileio
            Alias idisk
    Target iqn.2005-04.com.my:storage.disk2.sys2.icca
            User iscsiuser 1234567890abc
            Lun 3 /dev/sda7 fileio
            Alias icca
    说明:password 长度必须不小于12个字符, Alias是别名, 不知为何这个别名在
       Client端显示不出来.
  分区:我只有一个SCSI盘,所以:
        /dev/sda3: Share storage,容量越大越好
        /dev/sda5: raw1, 建Cluster要的rawdevice, 我给了900M
        /dev/sda6: raw2, 建Cluster要的rawdevice, 我给了900M
        /dev/sda7: cca, 建GFS要的,我给了64M
        (/dev/sda4是Extended分区,在其中建了sda5,6,7)
  #Reboot,用service iscsi-target start启ISCSI server(我觉得比建议的好,可以
    用service iscsi-target status看状态)
2.Setup ISCSI Client(on two real server)
  Server: PIII 1.4,512M, Dell 1650,Redhat AS3U4(用AS3U5更好),2.4.21-27.EL
  #vi /etc/iscsi.conf
     DiscoveryAddress=192.168.1.124
       OutgoingUsername=iscsiuser
       OutgoingPassword=1234567890abc
     Username=iscsiuser
     Password=1234567890abc
     LoginTimeout=15
     IncomingUsername=iscsiuser
     IncomingPassword=1234567890abc
     SendAsyncTest=yes
  #service iscsi restart
  #iscsi-ls -l
   ..., 精简如下:
    /dev/sdb:iraw2
    /dev/sdc:iraw1
    /dev/sdd:idisk
    /dev/sde:icca
  注意: 在real server中ISCSI device的顺序很重要,两个real server中一定要一样,如不一样
     就改ISCSI Server中的设置,多试几次
3.Install Redhat Cluster suite
  先下载Cluster Suite的ISO, AS3的我是从ChinaUnix.net找到的下载点, 安装clumanager和
  redhat-config-cluster。没有Cluster Suite的ISO也没关系,从
  ftp://ftp.redhat.com/pub/redhat/linux/updates/enterprise/3ES/en/RHCS/SRPMS/下载
  clumanager-1.2.xx.src.rpm,redhat-config-cluster-1.0.x.src.rpm,编译后安装,应该更好:
    #rpm -Uvh clumanager-1.2.26.1-1.src.rpm
    #rpmbuild -bs /usr/src/redhat/SPECS/clumanager.spec
    #rpmbuild --rebuild --target i686 /usr/src/redhat/SRPMS/clumanager-1.2.26.1-1.src.rpm
    还有redhat-config-cluster-1.0.x.src.rpm,也装好
4.Setup Cluster as HA module
  详细步骤我就不写了,网上有很多文章,我也是看了别人的文章学会的,不过人家是用VMWARE,
  而我是用真的机子+ISCSI,raw device就是/dev/sdb,/dev/sdc, 然后就
  mount /dev/sdd /u01, mkfs.ext3 /u01 ......
  设好后会发现ISCSI有问题:同时只能有一个Client联接写盘,如果
  两个Client同时联ISCSI的Share Storge,一个Client写,另一个Client是看不到的,而且此时文
  件系统已经破坏了,Client重联ISCSI时会发现文件是坏的,用fsck也修复不了。
ISCSI真的是鸡肋吗?
 NO!从GOOGLE上我终于查到ISCSI只有用Cluster File System才能真正用于Share Storage!
  而Redhat买下的GFS就是一个!
5.Setup GFS on ISCSI
  GFS只有Fedora Core4才自带了,而GFS又一定要用到Cluster Suite产生的/etc/cluster.xml文件,
  我没见FC4有Cluster Suite,真不知Redhat给FC4带GFS干嘛,馋人吗?
  好,闲话少说,下载:c处的GFS-6.0.2.20-2.src.rpm, 按a处的gfs.txt编译安装,不过关于
  cluster.ccs,fence.ccs,nodes.ccs的设置没说,看b的文档,我总算弄出来了,都存在
  /root/cluster下,存在别的地方也行,不过我不知道有没有错,我没有光纤卡,文档又没讲ISCSI
  的例子,不过GFS能启动的。
  #cat cluster.ccs 
      cluster {
              name = "Cluster_1"
              lock_gulm {
                  servers = ["cluster1", "cluster2"]
                  heartbeat_rate = 0.9
                  allowed_misses = 10
              }
      }
  注:name就是Cluster Suite设置的Cluster name, servers就是Cluster member的Hostname,别忘
  了加进/etc/hosts;allowed_misses我开始设为1,结果跑二天GFS就会死掉,改为10就没死过了。
  #cat fence.ccs 
      fence_devices{
              admin {
                    agent = "fence_manual"
              }
      }
  #cat nodes.ccs 
      nodes {
         cluster1 {
            ip_interfaces {
               hsi0 = "192.168.0.1"
            }
            fence {
               human {
                  admin {
                     ipaddr = "192.168.0.1"
                  }
               }
            }
         }
         cluster2 {
            ip_interfaces {
               hsi0 = "192.168.0.2"
            }
            fence {
               human {
                  admin {
                     ipaddr = "192.168.0.2"
                  }
               }
            }
         }
      }
    注:ip就是心跳线的ip
    这三个文件建在/root/cluster下,先建立Cluster Configuration System:
    a.#vi /etc/gfs/pool0.cfg
      poolname pool0
      minor 1 subpools 1
      subpool 0 8 1 gfs_data
      pooldevice 0 0 /dev/sde1
    b.#pool_assemble -a pool0
    c.#ccs_tool create /root/cluster /dev/pool/pool0
    d.#vi /etc/sysconfig/gfs
      CCS_ARCHIVE="/dev/pool/pool0"
      
    再Creating a Pool Volume,就是我们要的共享磁盘啦,
    a.#vi /etc/gfs/pool1.cfg
      poolname pool1
      minor 2 subpools 1
      subpool 0 128 1 gfs_data
      pooldevice 0 0 /dev/sdd1
    b.#pool_assemble -a pool1
    c.#gfs_mkfs -p lock_gulm -t Cluster_1:gfs1 -j 8 /dev/pool/pool1
    d.#mount -t gfs -o noatime /dev/pool/pool1 /u01
    下面是个GFS的启动脚本,注意real1和real2必须同时启动lock_gulmd进程,第一台lock_gulmd
    会成为Server并等Client的lock_gulmd,几十秒后没有响应会fail,GFS启动失败。Redhat建议
    GFS盘不要写进/etc/fstab。
    #cat gfstart.sh
      #!/bin/sh
      depmod -a
      modprobe pool
      modprobe lock_gulm
      modprobe gfs
      sleep 5
      service iscsi start
      sleep 20
      service rawdevices restart
      pool_assemble -a pool0
      pool_assemble -a pool1
      service ccsd start
      service lock_gulmd start
      mount -t gfs /dev/pool/pool1 /s02 -o noatime
      service gfs status 
6. Setup Linux LVS      
   LVS是章文嵩博士发起和领导的优秀的集群解决方案,许多商业的集群产品,比如RedHat的
   Piranha,Turbolinux公司的Turbo Cluster等,都是基于LVS的核心代码的。
   我的系统是Redhat AS3U4,就用Piranha了。
   从rhel-3-u5-rhcs-i386.iso安装piranha-0.7.10-2.i386.rpm,ipvsadm-1.21-9.ipvs108.i386.rpm
   (http://distro.ibiblio.org/pub/linux/distributions/caoslinux/centos/3.1/contrib/i386/RPMS/)
   装完后service httpd start & service piranha-gui start,就可以从http://xx.xx.xx.xx:3636管理或
   设置了,当然了,手工改/etc/sysconfig/ha/lvs.cf也一样。
   #cat /etc/sysconfig/ha/lvs.cf
     serial_no = 80
     primary = 10.3.1.101
     service = lvs
     rsh_command = ssh
     backup_active = 0
     backup = 0.0.0.0
     heartbeat = 1
     heartbeat_port = 1050
     keepalive = 6
     deadtime = 18
     network = nat
     nat_router = 192.168.1.1 eth1:1
     nat_nmask = 255.255.255.0
     reservation_conflict_action = preempt
     debug_level = NONE
     virtual lvs1 {
          active = 1
          address = 10.3.1.254 eth0:1
          vip_nmask = 255.255.255.0
          fwmark = 100
          port = 80
          persistent = 60
          pmask = 255.255.255.255
          send = "GET / HTTP/1.0\r\n\r\n"
          expect = "HTTP"
          load_monitor = ruptime
          scheduler = wlc
          protocol = tcp
          timeout = 6
          reentry = 15
          quiesce_server = 1
          server Real1 {
              address = 192.168.1.68
              active = 1
              weight = 1
          }
          server Real2 {
              address = 192.168.1.67
              active = 1
              weight = 1
          }
     }
     virtual lvs2 {
          active = 1
          address = 10.3.1.254 eth0:1
          vip_nmask = 255.255.255.0
          port = 21
          send = "\n"
          use_regex = 0
          load_monitor = ruptime
          scheduler = wlc
          protocol = tcp
          timeout = 6
          reentry = 15
          quiesce_server = 0
          server ftp1 {
              address = 192.168.1.68
              active = 1
              weight = 1
          }
          server ftp2 {
              address = 192.168.1.67
              active = 1
              weight = 1
          }
     }
   设置完后service pulse start, 别忘了把相关的client加进/etc/hosts
   #iptables -t mangle -A PREROUTING -p tcp -d 10.3.1.254/32 --dport 80 -j MARK --set-mark 100
   #iptables -t mangle -A PREROUTING -p tcp -d 10.3.1.254/32 --dport 443 -j MARK --set-mark 100
   #iptables -A POSTROUTING -t nat -p tcp -s 10.3.1.0/24 --sport 20 -j MASQUERADE
   运行以上三行命令并存入/etc/rc.d/rc.local,用ipvsadm看状态:
   #ipvsadm                                                                
   IP Virtual Server version 1.0.8 (size=65536)                          
   Prot LocalAddress:Port Scheduler Flags                                
     -> RemoteAddress:Port           Forward Weight ActiveConn InActConn 
   TCP  10.3.1.254:ftp wlc                                            
     -> cluster2:ftp                  Masq    1      0          0         
     -> cluster1:ftp                  Masq    1      0          0         
   FWM  100 wlc persistent 60                                                
     -> cluster1:0                    Masq    1      0          0         
     -> cluster2:0                    Masq    1      0          0  
   注意:a.Firewall Mark可以不要,我反正是加了,文档说有https的话加上,值我选了100,
         b.Virtual IP别加进/etc/hosts,我上过当,80端口时有时无的,
         c.eth0:1,eth1:1是piranha产生的,别自己手工设置,我干过这画蛇添足的事,网上有
           些帖子没说清,最后是看Redhat的文档才弄清楚的。
         d.The LVS router can monitor the load on the various real servers by using 
           either rup or ruptime. If you select rup  from the drop-down menu, each real
           server must run the rstatd service. If you select ruptime, each real server 
           must run the rwhod service.Redhat的原话,就是如选rup的监控模式real server上
           都要运行rstatd进程,如选ruptime就要运行rwhod进程。
         e.Real Server同Router相联的网卡的Gateway必须是Router的那块网卡的VIP,举本例:
           Router的eth1同两个real server的eth0相联,如VIP eth1:1=192.168.1.1,则real 
           server 的eth0的Gateway=192.168.1.1
7.Setup TOMCAT5.59+JDK1.5(用Redhat自带的Apache)
  a.#tar xzvf jakarta-tomcat-5.5.9.tar.gz
    #mv jakarta-tomcat-5.5.9 /usr/local
    #ln -s /usr/local/jakarta-tomcat-5.5.9 /usr/local/tomcat
  b.#jdk-1_5_0_04-linux-i586.bin
    #mv jdk1.5.0_4 /usr/java
    #ln -s /usr/java/jdk1.5.0_4 /usr/java/jdk
  c.#vi /etc/profile.d/tomcat.sh
     export CATALINA_HOME=/usr/local/tomcat
     export TOMCAT_HOME=/usr/local/tomcat
  d.#vi /etc/profile.d/jdk.sh
     if ! echo ${PATH} | grep "/usr/java/jdk/bin" ; then
       JAVA_HOME=/usr/java/jdk
       export JAVA_HOME
       export PATH=/usr/java/jdk/bin:${PATH}
       export CLASSPATH=$JAVA_HOME/lib
     fi
  e.#chmod 755 /etc/profile.d/*.sh
  f.重新用root登录,让tomcat.sh和jdk.sh起作用,
    #tar xzvf jakarta-tomcat-connectors-jk2-src-current.tar.gz
    #cd jakarta-tomcat-connectors-jk2-2.0.4-src/jk/native2/ 
    #./configure --with-apxs2=/usr/sbin/apxs --with-jni --with-apr-lib=/usr/lib
    #make
    #libtool --finish /usr/lib/httpd/modules
    #cp ../build/jk2/apache2/mod_jk2.so ../build/jk2/apache2/libjkjni.so /usr/lib/httpd/modules/ 
  g.#vi /usr/local/tomcat/bin/catalina.sh
    在# Only set CATALINA_HOME if not already set后加上以下两行:
    serverRoot=/etc/httpd
    export serverRoot 
  h.#vi /usr/local/tomcat/conf/jk2.properties
    serverRoot=/etc/httpd                           
    apr.NativeSo=/usr/lib/httpd/modules/libjkjni.so 
    apr.jniModeSo=/usr/lib/httpd/modules/mod_jk2.so
  i.#vi /usr/local/tomcat/conf/server.xml,
    在</Host>前加上以下几行,建了两个VirtualPath:myjsp和local,一个指向share storage,
    一个指向real server本地
    <Context path="/myjsp" docBase="/u01/www/myjsp" debug="0"/>
      <Logger className="org.apache.catalina.logger.FileLogger" directory="/var/log/httpd"
       prefix="cluster.log." suffix=".txt" timestamp="true" />
      <Valve className="org.apache.catalina.valves.AccessLogValve" directory="/var/log/httpd"
       prefix="cluster_access.log." suffix=".txt" pattern="common" resolveHosts="false" />
    <Context path="/local" docBase="/var/www/html" debug="1" reloadable="true" crossContext="true" />
  j.#vi /etc/httpd/conf/workers2.properties
    #[logger.apache2]
    #level=DEBUG
    [shm]
      file=/var/log/httpd/shm.file
      size=1048576
    [channel.socket:localhost:8009]
      tomcatId=localhost:8009
      keepalive=1
      info=Ajp13 forwarding over socket
    [ajp13:localhost:8009]
      channel=channel.socket:localhost:8009
    [status:status]
      info=Status worker, displays runtime informations
    [uri:/*.jsp]
      worker=ajp13:localhost:8009
      context=/   
  k.#vi /etc/httpd/conf/httpd.conf
    改:DocumentRoot "/u01/www"  
    加:
    在LoadModule最后加:
    LoadModule jk2_module modules/mod_jk2.so
    JkSet config.file /etc/httpd/conf/workers2.properties
    在#<VirtualHost *>之前加:
    <Directory ~ "/WEB-INF/">
     Order allow,deny
     Deny from all
    </Directory> 
  l:#mkdir /u01/ftproot
    #mkdir /u01/www
    #mkdir /u01/www/myjsp
  m:在每个real server上生成index.jsp
    #vi /var/www/html/index.jsp
    <%@ page import="java.util.*,java.sql.*,java.text.*" contentType="text/html"
    %>
    <%
    out.println("test page on real server 1");
    %>     
    在real server2上就是"test page on real server 2"
  n:下载jdbc Driver
    http://www.oracle.com/technology/software/tech/java/sqlj_jdbc/htdocs/jdbc9201.html  
    可惜只有for JDK1.4的,在两台real server上分别
    #cp -R /usr/local/tomcat/webapps/webdav/WEB-INF /u01/www/myjsp
    #cp ojdbc14.jar ojdbc14_g.jar ocrs12.zip /u01/www/myjsp/WEB-INF/lib
  o:假设我有一台OracleServer,ip=10.3.1.211,sid=MYID,username=my,password=1234,并有Oracle
    的例子employees的read权限,或干脆把这个table拷过来,我是Oracle9i中的
    #vi /u01/www/myjsp/testoracle.jsp
    <%@ page contentType="text/html" %>
    <%@ page import="java.sql.*"%>
    <?xml version="1.0"?>
    <html">
    <head>
    <meta http-equiv="Content-Type" content="text/html" />
    <title>Test ORACLE Employees</title>
    </head>
    
    <body>
    <%
           String OracleDBDriver="oracle.jdbc.driver.OracleDriver";
           String DBUrl="jdbc:oracle:thin:@10.3.1.211:1521:MYID";
           String UserID="my";
           String UserPWD="1234";
          
           Connection conn=null;
           Statement  stmt=null;
           ResultSet  rs=null;
           try
           {
           Class.forName(OracleDBDriver);
           }
           catch(ClassNotFoundException ex)
           {     
            System.out.println("Class.forname:"+ex);
            }
        conn=DriverManager.getConnection(DBUrl,UserID,UserPWD);
        stmt=conn.createStatement();
        String sql="select * from EMPLOYEES";
        rs = stmt.executeQuery(sql);
       
            out.print("<table border>");
              out.print("<tr>");
                out.print("<th width=100>"+"EMPLOYEE_ID");
                   out.print("<th width=50>"+"FIRST_NAME");
                    out.print("<th width=50>"+"LAST_NAME");
                    out.print("<th width=50>"+"EMAIL");
                    out.print("<th width=50>"+"PHONE_NUMBER");
                    out.print("<th width=50>"+"HIRE_DATE");
                    out.print("<th width=50>"+"JOB_ID");
    
              out.print("<tr>");
     try
       {
            while(rs.next())
            {
              out.print("<tr>");
                int n=rs.getInt(1);
                   out.print("<td>"+n+"</td>");
                   String e=rs.getString(2);
                   out.print("<td>"+e+"</td>");
                   //String e=rs.getString(3);
                   out.print("<td>"+rs.getString(3)+"</td>");
                   out.print("<td>"+rs.getString(4)+"</td>");
                   out.print("<td>"+rs.getString(5)+"</td>");
                   out.print("<td>"+rs.getString(6)+"</td>");
                   out.print("<td>"+rs.getString(7)+"</td>");
    
              out.print("</tr>");
            }
        }
           catch(SQLException ex)
           {
                  System.err.println("ConnDB.Main:"+ex.getMessage());
           }
    
            out.print("</table>"); 
            rs.close();
            stmt.close();
            conn.close(); 
    %> 
    </body>
    </html>
   p:#vi /u01/www/index.html
    <p>
    <META HTTP-EQUIV="Refresh" CONTENT="10; URL=http://10.3.1.254/myjsp/testoracle.jsp">
    </BODY>
    <p>
    <a href="http://10.3.1.254/local/index.jsp">WEB Local</a>
    <p>
    <a href="http://10.3.1.254/myjsp/testoracle.jsp">Test Oracle WEB</a>
    </HTML>
   
   q:在两台real server上分别
     #vi /usr/local/tomcat/conf/tomcat-users.xml
     加下面一行,允许页面管理:
     <user username="manager" password="tomcat" roles="manager"/>
   r:在两台real server上分别
     #service httpd restart           
     #/usr/local/tomcat/bin/startup.sh
   s:打开http://1092.168.1.68:8080和http://1092.168.1.67:8080,选Tomcat Manager,用
     manager/tomcat登录,虚拟目录/myjsp和/local应该Start了
   在两台机子上分别打开网页http://10.3.1.254,选WEB Local,可以看到一台显示:
   "test page on real server 1",另一台为"test page on real server 2",同时在Router上
   ipvsadm可以看到每个real server的联接数
8.设置FTP服务
  #vi /etc/vsftpd/vsftp.conf,在两台real server上分别加入以下几行:
  anon_root=/u01/ftproot
  local_root=/u01/ftproot
  setproctitle_enable=YES
  #service vsftpd start
现在LVM+GFS+ISCSI+TOMCAT就设置好了,我们可以用Apache Jmeter来测试LVM的性能,两台机子上
分别运行jmeter,都指向10.3.1.254/myjsp/testoracle.jsp,各200个threads同时运行,在Router
上用ipvsadm可以监控,Oracle Server的性能可要好,否则大量的http进程会hang在real server
上,ipvsadm也会显示有个real server失去了。测试时real server的CPU idle会降到70%,而Router
的CPU idle几乎不动

你可能感兴趣的:(oracle,tomcat,linux,redhat,SQL Server)