创新互联公司坚持“要么做到，要么别承诺”的工作理念，服务领域包括：网站设计、网站建设、企业官网、英文网站、手机端网站、网站推广等服务，满足客户于互联网时代的西湖网站设计、移动媒体设计的需求，帮助企业找到有效的互联网解决方案。努力成为您成熟可靠的网络建设合作伙伴！
HA(High Available)即由两台计算机组成并对外提供一组相同的服务，也叫做一主一备模式。正常情况下由主服务器提供服务，备服务器处于待机备用，备机可以随时接替主服务器的工作。也就是当主服务器宕机或所提供的服务不可用时，备用服务器会主动激活并且替换主服务器继续提供服务，这时主服务器上的服务资源包括网络(IP)、存储、服务(Web/数据库)就会转移到备机接管，从而提供不间断的服务。基于此，便可以将服务器的宕机时间减少到最低，对业务实现不中断或短暂中断。
由一组计算机(多台)组成一个整体并向用户提供相同的网络资源或服务，这种模式叫做HA集群(High Available Cluster)。
Oracle HA高可用，又叫做双机热备，一般用于关键性业务。
Oracle双机热备模式
常用的有Data Guard、RAC(Real Application Clusters)、基于HA软件实现的双机热备。
l Data Guard采用重做日志复制技术，对主业务数据进行实时的异步同步备份，有一主一备、一主多备模式，一般常用于容灾(异地备份、灾难恢复)；存储独立、数据完整备份；备机可以做数据分析、报表统计等；主备自由切换，能实现自动快速故障转移；备用角色强制转换为主用角色会破环Data Guard架构；设计复杂、维护难度大
l RAC多个实例同时运行，无主备概念，有集群负载功能，其中某一台down机，不影响整体服务，不存在故障切换时间，可以提供高性能服务；存储共享，由ASM管理存储；硬件成本低；设计简单、维护难度大
l HA双机热备可以提供高可用性，保证业务的持续稳定运行，可以实现自动快速故障转移，存在短暂的切换时间(10-30s)。一般用于关键性业务；存储共享；有商业的HA方案，也可以使用开源的高可用软件keepalved、heartbeat搭建实现HA方案；设计复杂、维护难度小
l HA双机和Data Guard
l Data Guard和RAC
l Oracle HA双机热备+Keepalived
l Oracle HA双机+DataGuard+Keepalived集群
软件版本：
Orace 11g R2
Keepalived 1.3.2
主服务器: DB1
备服务器: DB2
Hostname | IP | OS | Role |
hmdg-db1 | 172.16.10.25 | CentOS6.9 | MASTER |
hmdg-db2 | 172.16.10.26 | CentOS6.9 | BACKUP |
VIP | 172.16.10.130、172.16.10.131 | ||
Share Disk | /dev/sdb1 mount on /oradata | ||
ORACLE_BASE | /u01/app/oracle | ||
ORACLE_HOME | /u01/app/oracle/product/11.2.0/db_1 | ||
ORACLE_SID | HMODB | ||
Datadir | /oradata/HMODB | ||
Controlfiles | /oradata/HMODB/control01.ctl, /u01/app/oracle/flash_recovery_area/HMODB/control02.ctl, /home/oracle/rman/HMODB/control03.ctl | ||
1、 同时在主备上安装Oracle数据库软件，数据库的安装目录和环境变量保持一致
2、 仅在主服务器上建立数据库实例(主服务器上挂载共享存储并使用DBCA新建库)
3、 主服务创建数据库实例后，将主库上的控制文件、参数文件(SPFILE)、以及密码文件传输到备用服务器上相应的目录
l 备用服务器DB2
##创建对应的数据库文件目录
[root@hmdg-db2 ~]# su - oracle
[oracle@hmdg-db2 ~]$ mkdir /u01/app/oracle/admin/HMODB
[oracle@hmdg-db2 ~]$ mkdir /u01/app/oracle/admin/HMODB/{adump,dpdump,pfile}
[oracle@hmdg-db2 ~]$ mkdir /u01/app/oracle/flash_recovery_area/HMODB
##密码文件
[oracle@hmdg-db2 ~]$ scp oracle@hmdg-db1:/u01/app/oracle/product/11.2.0/db_1/dbs/orapwHMODB /u01/app/oracle/product/11.2.0/db_1/dbs/orapwHMODB
##SPFILE参数文件
[oracle@hmdg-db2 ~]$ scp oracle@hmdg-db1:/u01/app/oracle/product/11.2.0/db_1/dbs/spfileHMODB.ora /u01/app/oracle/product/11.2.0/db_1/dbs/spfileHMODB.ora
##控制文件，仅需要将位于非共享存储的数据目录下的控制文件传输到其他磁盘或目录位置下
[oracle@hmdg-db2 ~]$ scp oracle@hmdg-db1:/u01/app/oracle/flash_recovery_area/HMODB/control02.ctl /u01/app/oracle/flash_recovery_area/HMODB/control02.ctl
[oracle@hmdg-db2 ~]$cp /u01/app/oracle/flash_recovery_area/HMODB/control02.ctl /home/oracle/rman/HMODB/control03.ctl1、 主服务器上关闭数据库实例
2、 主服务器上卸载共享存储
3、 备服务器上挂载共享存储
4、 备服务器上启动监听
5、 备服务器上启动数据库实例
注意：不要同时挂载共享存储启动数据库实例，否则会导致数据不一致
l 主服务器DB1
##关闭数据库实例 [oracle@hmdg-db1 ~]$ sqlplus / as sysdba SQL> shutdown immediate ##关闭数据库监听 [oracle@hmdg-db1 ~]$ lsnrctl stop ##卸载共享存储 [oracle@hmdg-db1 ~]$ umount /oradata
l 备服务器DB2
##挂载/dev/sdb1共享存储到/oradta下 [root@hmdg-db2 ~]# mount /dev/sdb1 /oradata/ ##启动数据库监听 [oracle@hmdg-db2 ~]$ lsnrctl start [oracle@hmdg-db2 ~]$ export ORACLE_SID=HMODB ##启动数据库实例 [oracle@hmdg-db2 ~]$ sqlplus / as sysdba SQL> startup ORACLE instance started. Total System Global Area 1603411968 bytes Fixed Size 2213776 bytes Variable Size 1056966768 bytes Database Buffers 536870912 bytes Redo Buffers 7360512 bytes Database mounted. Database opened. SQL> ##检验数据库 SQL> SHOW PARAMETER DB_NAME NAME TYPE VALUE ------------------------------------ ----------- ------------------------------ db_name string HMODB SQL> SELECT OPEN_MODE,NAME,DATABASE_ROLE FROM V$DATABASE; OPEN_MODE NAME DATABASE_ROLE -------------------- --------- ---------------- READ WRITE HMODB PRIMARY
由于安装数据库实例时，创建了多个控制文件，并且这些控制文件都保存在不同的磁盘上，在主备进行切换的时，位于非共享存储的控制文件需要手动进行同步。这是由于在关闭数据库实例时Oracle同时将当前数据库状态信息记录到控制文件中，在没有启动数据库实例的备机就无法同步自身的其他的控制文件。这时就需要使用共享存储里的控制文件进行恢复。
1、 在DB2关闭正在运行数据库实例
##DB2: [oracle@hmdg-db2 ~]$ sqlplus / as sysdba SQL> shutdown immediate [root@hmdg-db2 ~]# umount /oradata/
2、在DB1上进行恢复启动数据库实例
##DB1： [root@hmdg-db1 ~]# mount /dev/sdb1 /oradata/ [oracle@hmdg-db1 ~]$ export ORACLE_SID=HMODB [oracle@hmdg-db1 ~]$ sqlpus / as sysdba SQL> startup nomount SQL> alter database mount; SQL> alter database open;
如果控制文件在不同的磁盘目录，那么就需要执行这一步恢复控制文件
[oracle@hmdg-db1 ~]$ rman target / RMAN> restore controlfile from '/oradata/HMODB/control01.ctl';
启动数据库监听
[oracle@hmdg-db1 ~]$ lsnrctl start
或者直接只有操作系统的cp、scp等命令进行拷贝恢复
[oracle@hmdg-db1 ~]$ cp /oradata/HMODB/control01.ctl /u01/app/oracle/flash_recovery_area/HMODB/control02.ctl SQL> SELECT NAME,OPEN_MODE,NAME,DATABASE_ROLE FROM V$DATABASE; NAME OPEN_MODE NAME DATABASE_ROLE --------- -------------------- --------- ---------------- HMODB READ WRITE HMODB PRIMARY
配置oracle用户的ssh密钥，用于主备通过oracle身份自动检测对方的数据共享存储状态。在keepalived脚本管理中自动检查对方的挂载状态，并在主备切换的过程中，待接管的服务器等待对方umount共享存储之后，正常启动数据库实例。（如果主服务异常关机或网络中断，不需要等待对方umount存储，主服务器异常关机时keepalived会即时检测到并自动切换到备机上）
[oracle@hmdg-db1 ~]$ ssh-keygen -t rsa -b 2048 [oracle@hmdg-db1 ~]$ ssh-copy-id -i ~/.ssh/id_rsa.pub oracle@hmdg-db2 [oracle@hmdg-db2 ~]$ ssh-keygen -t rsa -b 2048 [oracle@hmdg-db2 ~]$ ssh-copy-id -i ~/.ssh/id_rsa.pub oracle@hmdg-db1
添加主备防火墙互通规则
# iptables -I INPUT -p tcp -s 172.16.10.0/24 -j ACCEPT # iptables -I INPUT -p udp -s 172.16.10.0/24 -j ACCEPT
主服务器的keepalived配置，备用服务器需要做相应的修改
! Configuration File for keepalived
global_defs {
notification_email {
mail@huangming.org
}
notification_email_from Alexandre.Cassen@firewall.loc
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id ORACLE_HA_MASTER #backup为ORACLE_HA_BACKUP
!vrrp_skip_check_adv_addr
}
vrrp_sync_group ORACLE_HA_GROUP {
group {
Oracle_HA_1 #定义一个vrrp实例组
}
}
vrrp_script monitor_oracle_status { #oracle数据库服务器状态检查脚本
script "/usr/local/keepalived/scripts/monitor.sh"
interval 10 #脚本执行时间间隔
fall 3 #脚本执行的最大失败次数
rise 1 #脚本执行成功一次则返回正常状态
weight 20 #脚本执行成功时的附加权值
}
vrrp_script change_monitor_status { #服务器状态监控并修改主备检查脚本
script "/usr/local/keepalived/scripts/change_monitor_status.sh"
interval 2
}
vrrp_instance Oracle_HA_1 {
state BACKUP #主备都设置为BACKUP
interface eth2 #绑定vip的网络接口
virtual_router_id 200 #主备保存一致
priority 100 #备为90
advert_int 2 #主备状态同步时间间隔
nopreempt #主设置为不抢占模式
authentication {
auth_type PASS
auth_pass 11112222
}
virtual_ipaddress {
172.16.10.130/24 dev eth2 #需要绑定的VIP
172.16.10.131/24 dev eth2
}
track_script {
monitor_oracle_status
change_monitor_status
}
track_interface {
eth0 #同时监控检测其他的网接口
}
notify_master "/usr/local/keepalived/scripts/keepalived_notify.sh master"
notify_backup "/usr/local/keepalived/scripts/keepalived_notify.sh backup"
notify_fault "/usr/local/keepalived/scripts/keepalived_notify.sh fault"
notify_stop "/usr/local/keepalived/scripts/keepalived_notify.sh stop"
! notify管理脚本的定义
! 分别对应keepalived转换为master、backup、fault、stop状态时执行的脚本
! 在oracle的双机管理中主要是依赖这些脚本实现资源的接管
}[root@hmdg-db1 scripts]# ls -l /usr/local/keepalived/scripts/ -rwxr-xr-x 1 root root 818 Apr 14 20:13 change_monitor_status.sh -rw-r--r-- 1 root root 57 Apr 14 20:13 Controlfile.sql -rw-r--r-- 1 root root 1308 Apr 14 20:13 keepalived -rwxr-xr-x 1 root root 10549 Apr 14 23:18 keepalived_notify.sh -rwxr-xr-x 1 root root 926 Apr 14 23:12 monitor_primary_oracle.sh -rwxr-xr-x 1 root root 895 Apr 15 16:46 monitor.sh -rwxr-xr-x 1 root root 895 Apr 14 20:13 monitor_standby_oracle.sh -rw-r--r-- 1 root root 1657 Apr 15 15:13 oracle_init.sh -rwxr-xr-x 1 root root 8441 Apr 14 20:13 notify.sh_20180413
关于这些keepalived脚本，由于篇幅有限我没有全部贴到本文中，大家可以访问我的Github仓库找到：https://github.com/hmlinux/oracle-keepalived
l keepalived_notify.sh #keepalived的notify脚本，用于实现主备的资源自动切换
l monitor.sh #keepalved的监控脚本，监控oracle数据库运行状态、共享存储状态
l change_monitor_status.sh #异常或正常检测脚本，当备服务器处于正常运行状态，并且处于keepalived的BACKUP状态时，将自身的监控脚本修改为在备机就绪时执行都返回正常结果，主要作用时让正常的备机可以随时接管主服务器的资源
l monitor_primary_oracle.sh #处于MASTER状态时的服务器监控脚本
l monitor_standby_oracle.sh #处于BACKUP状态时的服务器监控脚本
l oracle_init.sh #notify脚本调用oracle启动、停止、状态检测命令
keepalived_notify.sh
#!/bin/bash
# author: hm@huangming.org
# keepalived notify script
#bind vip interface
interface="eth2"
#keepalived virtual_ipaddress
virtual_ipaddress="172.16.10.130","172.16.10.131"
MASTER_HOSTNAME="hmdg-db1" #DB1
BACKUP_HOSTNAME="hmdg-db2" #DB2
LOGDIR="/usr/local/keepalived/log"
LOGFILE="$LOGDIR/keepalived_haswitch.log"
TMPLOG=/tmp/notify_.log
sharedisk="/dev/sdb1"
sharedisk_mount_point="/oradata"
# oracle_datadir="/oradata/path/HMODB"
oracle_datadir="/oradata"
# Source oracle database instance startup/shutdown script
. /usr/local/keepalived/scripts/oracle_init.sh
control_files="/$oracle_datadir/$ORACLE_SID/control01.ctl","/u01/app/oracle/flash_recovery_area/$ORACLE_SID/control02.ctl","/home/oracle/rman/$ORACLE_SID/control03.ctl"
RETVAL=0
OLD_IFS="$IFS"
IFS=","
[ -d $LOGDIR ] || mkdir $LOGDIR
controlfile_backpath="/backup/oracle/control"
controlfile_back=$controlfile_backpath/control_$(date '+%Y%d%m%H%M%S')
[ -d $controlfile_backpath ] || mkdir -p $controlfile_backpath && chown -R oracle:oinstall $controlfile_backpath
info_log() {
printf "$(date '+%b %d %T %a') $HOSTNAME [keepalived_notify]: $1"
}
control01_ctl=`printf ${control_files[0]}`
backup_controlfile() {
su - oracle << EOF
export ORACLE_SID=$ORACLE_SID
$ORACLE_HOME/bin/sqlplus -S "/ as sysdba"
alter database backup controlfile to '$controlfile_back';
exit
EOF
}
#runuser -l oracle -c "export ORACLE_SID=$ORACLE_SID;rman target / cmdfile=/usr/scripts/ControlfileRestore.sql"
restore_controlfile() {
su - oracle << EOF
export ORACLE_SID=$ORACLE_SID
$ORACLE_HOME/bin/rman target / nocatalog
RESTORE CONTROLFILE FROM '$control01_ctl';
exit
EOF
}
ssh_p=$(netstat -ntp | awk '/sshd/{print $4}' | awk -F':' '{print $2}' | head -1)
chk_remote_node_sharedisk() {
if [ $(hostname) == $MASTER_HOSTNAME ];then
if ping -c1 -w2 $BACKUP_HOSTNAME &>/dev/null;then
runuser -l oracle -c "ssh -p$ssh_p $BACKUP_HOSTNAME 2>/dev/null df | grep $sharedisk | wc -l"
fi
fi
if [ $(hostname) == $BACKUP_HOSTNAME ];then
if ping -c1 -w2 $MASTER_HOSTNAME &>/dev/null;then
runuser -l oracle -c "ssh -p$ssh_p $MASTER_HOSTNAME 2>/dev/null df | grep $sharedisk | wc -l"
fi
fi
}
master() {
info_log "Database Switchover To MASTER\n"
info_log "Check remote node sharedisk mounted.\n"
i=1
while (($i<=30))
do
chk_status=$(chk_remote_node_sharedisk)
if [ $chk_status -ge 1 ];then
info_log "$sharedisk is already mounted on remote node or busy. checking [$i]...\n"
else
info_log "$sharedisk check passed.\n"
break
fi
if [ $i -eq 20 ];then
info_log "Disk status abnormal.\n"
exit 1
fi
sleep 1
i=$(($i+1))
done
ismount=$(df -h | grep $sharedisk | grep $sharedisk_mount_point | wc -l)
if [ $ismount -eq 0 ];then
info_log "mount $sharedisk on $sharedisk_mount_point\n"
mount $sharedisk $sharedisk_mount_point
RETVAL=$?
if [ $RETVAL -eq 0 ];then
#shutdown_instance
info_log "restore controlfile 1\n"
startup_nomount
restore_controlfile
else
info_log "Error: $sharedisk cannot mount or $sharedisk_mount_point busy\n"
exit $RETVAL
fi
else
disk=$(df -h | grep $sharedisk_mount_point | awk '{print $1}')
if [ $disk == $sharedisk ];then
info_log "mount: $sharedisk is already mounted on $sharedisk_mount_point\n"
else
info_log "Warning: $sharedisk already mounted on $disk\n"
fi
fi
status=$(check_instance_status | grep -Eio -e "\bOPEN\b" -e "\bMOUNTED\b" -e "\bSTARTED\b")
if [ "$status" == "OPEN" ];then
info_log "a database already open by the instance.\n"
elif [ "$status" == "MOUNTED" ];then
info_log "re-open database instance\n"
open_instance | tee $TMPLOG
opened=$(cat $TMPLOG | grep -Eio "\bDatabase altered\b")
if [ "$opened" != "Database altered" ];then
info_log "Error: database instance open fail!\n"
exit 2
fi
elif [ "$status" == "STARTED" ];then
info_log "alter database to mount\n"
mount_instance | tee $TMPLOG
mounted=$(cat $TMPLOG | grep -Eio "\bDatabase altered\b")
if [ "$mounted" == "Database altered" ];then
info_log "alter database to open\n"
open_instance | tee $TMPLOG
opened=$(cat $TMPLOG | grep -Eio "\bDatabase altered\b")
if [ "$opened" == "Database altered" ];then
info_log "Database opened.\n"
else
info_log "Database open failed\n"
exit 4
fi
else
info_log "Database mount failed\n"
exit 3
fi
else
info_log "Startup database and open instance\n"
shutdown_instance &>/dev/null
startup_instance | tee $TMPLOG
started=$(cat $TMPLOG | grep -Eio "\bDatabase opened\b")
if [ "$started" != "Database opened" ];then
info_log "Database instance open fail.\n"
info_log "restore controlfile 2\n"
shutdown_instance | tee $TMPLOG
startup_nomount | tee $TMPLOG
restore_controlfile
shutdown_instance | tee $TMPLOG
startup_instance | tee $TMPLOG
started=$(cat $TMPLOG | grep -Eio "\bDatabase opened\b")
if [ "$started" != "Database opened" ];then
info_log "Database restore fail!\n"
exit 5
else
info_log "Database opened.\n"
fi
else
info_log "Database opened.\n"
fi
fi
info_log "Startup listener\n"
runuser -l oracle -c "lsnrctl status &>/dev/null"
if [ $? -eq 0 ];then
info_log "listener already started.\n"
else
info_log "starting listener...\n"
runuser -l oracle -c "lsnrctl start &>/dev/null"
if [ $? -eq 0 ];then
info_log "The listener startup successfully\n"
else
info_log "Listener start failure!\n"
fi
fi
echo
}
backup() {
info_log "Database Switchover To BACKUP\n"
ismount=$(df -h | grep $sharedisk | grep $sharedisk_mount_point | wc -l)
if [ $ismount -ge 1 ];then
disk=$(df -h | grep $sharedisk_mount_point | awk '{print $1}')
if [ $disk == $sharedisk ];then
status=$(check_instance_status | grep -Eio -e "\bOPEN\b" -e "\bMOUNTED\b" -e "\bSTARTED\b")
if [ "$status" == "OPEN" -o "$status" == "MOUNTED" ];then
info_log "Database instance state is mounted\n"
info_log "Backup current controlfile.\n"
echo -e "\nSQL> alter database backup controlfile to '$controlfile_back';\n"
backup_controlfile
info_log "Shutdown database instance, please wait...\n"
shutdown_instance | tee $TMPLOG
shuted=$(cat $TMPLOG | grep -Eio "\binstance shut down\b")
if [ "$shuted" == "instance shut down" ];then
info_log "Database instance shutdown successfully.\n"
else
info_log "Database instance shutdown failed.\n"
info_log "shutdown abort.\n"
shutdown_abort
fi
elif [ "$status" == "STARTED" ];then
info_log "Database instance state is STARTED\n"
info_log "Shutdown database instance, please wait...\n"
shutdown_instance | tee $TMPLOG
shuted=$(cat $TMPLOG | grep -Eio "\binstance shut down\b")
if [ "$shuted" == "instance shut down" ];then
info_log "Database instance shutdown successfully.\n"
else
info_log "Database instance shutdown failed.\n"
info_log "shutdown abort.\n"
shutdown_abort
fi
else
shutdown_instance | tee $TMPLOG
info_log "Database instance not available.\n"
fi
echo
info_log "umount sharedisk\n"
echo
umount $sharedisk_mount_point && RETVAL=$?
if [ $RETVAL -eq 0 ];then
info_log "umount $sharedisk_mount_point success.\n"
else
info_log "umount $sharedisk_mount_point fail!\n"
fi
else
info_log "$sharedisk is not mount on $sharedisk_mount_point or busy.\n"
fi
else
info_log "$sharedisk_mount_point is no mount\n"
fi
info_log "stopping listener...\n"
runuser -l oracle -c "lsnrctl status" &>/dev/null
RETVAL=$?
if [ $RETVAL -eq 0 ];then
runuser -l oracle -c "lsnrctl stop" &>/dev/null
RETVAL=$?
if [ $RETVAL -eq 0 ];then
info_log "The listener stop successfully\n"
else
info_log "Listener stop failure!\n"
fi
else
info_log "listener is not started.\n"
fi
echo
}
notify_master() {
echo -e "\n-------------------------------------------------------------------------------"
echo "`date '+%b %d %T %a'` $(hostname) [keepalived_notify]: Transition to $1 STATE";
echo "`date '+%b %d %T %a'` $(hostname) [keepalived_notify]: Setup the VIP on $interface $virtual_ipaddress";
}
notify_backup() {
echo -e "\n-------------------------------------------------------------------------------"
echo "`date '+%b %d %T %a'` $HOSTNAME [keepalived_notify]: Transition to $1 STATE";
echo "`date '+%b %d %T %a'` $HOSTNAME [keepalived_notify]: removing the VIP on $interface for $virtual_ipaddress";
}
case $1 in
master)
notify_master MASTER | tee -a $LOGFILE
master | tee -a $LOGFILE
;;
backup)
notify_backup BACKUP | tee -a $LOGFILE
backup | tee -a $LOGFILE
;;
fault)
notify_backup FAULT | tee -a $LOGFILE
backup | tee -a $LOGFILE
;;
stop)
notify_backup STOP | tee -a $LOGFILE
/etc/init.d/keepalived start
#sleep 6 && backup | tee -a $LOGFILE
;;
*)
echo "Usage: `basename $0` {master|backup|fault|stop}"
RETVAL=1
;;
esac
exit $RETVALl 将主服务器切换为BACKUP状态
执行./keepalived_notify.sh backup
[root@hmdg-db1 scripts]# ./keepalived_notify.sh backup ------------------------------------------------------------------------------- Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: Transition to BACKUP STATE Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: removing the VIP on eth2 for 172.16.10.130,172.16.10.131 Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: Database Switchover To BACKUP Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: /oradata is no mount Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: stopping listener... Apr 15 17:23:05 Sun hmdg-db1 [keepalived_notify]: listener is not started.
观察执行脚本的输出结果，这是由于在本机上并没有启动oracle实例，所以脚本最终提示本机并没有启动oracle数据库实例
l 将主服务器切换为MASTER状态
执行./keepalived_notify.sh master
[root@hmdg-db1 scripts]# ./keepalived_notify.sh master ------------------------------------------------------------------------------- Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: Transition to MASTER STATE Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: Setup the VIP on eth2 172.16.10.130,172.16.10.131 Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: Database Switchover To MASTER Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: Check remote node sharedisk mounted. Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 check passed. Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: mount /dev/sdb1 on /oradata Apr 15 17:23:43 Sun hmdg-db1 [keepalived_notify]: restore controlfile 1 ORACLE instance started. Total System Global Area 1603411968 bytes Fixed Size 2213776 bytes Variable Size 1056966768 bytes Database Buffers 536870912 bytes Redo Buffers 7360512 bytes Recovery Manager: Release 11.2.0.1.0 - Production on Sun Apr 15 17:23:44 2018 Copyright (c) 1982, 2009, Oracle and/or its affiliates. All rights reserved. connected to target database: HMODB (not mounted) using target database control file instead of recovery catalog RMAN> Starting restore at 15-APR-18 allocated channel: ORA_DISK_1 channel ORA_DISK_1: SID=572 device type=DISK channel ORA_DISK_1: copied control file copy output file name=/oradata/HMODB/control01.ctl output file name=/u01/app/oracle/flash_recovery_area/HMODB/control02.ctl output file name=/home/oracle/rman/HMODB/control03.ctl Finished restore at 15-APR-18 RMAN> Recovery Manager complete. Apr 15 17:23:46 Sun hmdg-db1 [keepalived_notify]: alter database to mount Database altered. Apr 15 17:23:50 Sun hmdg-db1 [keepalived_notify]: alter database to open Database altered. Apr 15 17:23:57 Sun hmdg-db1 [keepalived_notify]: Database opened. Apr 15 17:23:57 Sun hmdg-db1 [keepalived_notify]: Startup listener Apr 15 17:23:57 Sun hmdg-db1 [keepalived_notify]: starting listener... Apr 15 17:23:57 Sun hmdg-db1 [keepalived_notify]: The listener startup successfully
观察执行脚本的输出结果
1、脚本接收转换到MASTER信息
2、首先提示绑定的VIP
3、然后检测远程备用节点的磁盘是否挂载，或本机是否挂载了磁盘，检测成功则把磁盘挂载
4、启动数据库到nomount状态，并恢复控制文件
5、启动数据库实例到open状态
6、启动数据库监听，切换完成
注意：如果使用系统命令手动切换，应该要保证正在运行oracle服务的主节点先正常关闭，并且卸载共享存储之后，再启动备用节点的oracle服务(即不能同时挂载共享存储)
l 再次将主服务器切换到BACKUP状态
执行./keepalived_notify.sh backup
[root@hmdg-db1 scripts]# ./keepalived_notify.sh backup ------------------------------------------------------------------------------- Apr 15 17:34:54 Sun hmdg-db1 [keepalived_notify]: Transition to BACKUP STATE Apr 15 17:34:54 Sun hmdg-db1 [keepalived_notify]: removing the VIP on eth2 for 172.16.10.130,172.16.10.131 Apr 15 17:34:54 Sun hmdg-db1 [keepalived_notify]: Database Switchover To BACKUP Apr 15 17:34:55 Sun hmdg-db1 [keepalived_notify]: Database instance state is mounted Apr 15 17:34:55 Sun hmdg-db1 [keepalived_notify]: Backup current controlfile. SQL> alter database backup controlfile to '/backup/oracle/control/control_20181504173454'; Database altered. Apr 15 17:34:55 Sun hmdg-db1 [keepalived_notify]: Shutdown database instance, please wait... Database closed. Database dismounted. ORACLE instance shut down. Apr 15 17:35:04 Sun hmdg-db1 [keepalived_notify]: Database instance shutdown successfully. Apr 15 17:35:04 Sun hmdg-db1 [keepalived_notify]: umount sharedisk Apr 15 17:35:05 Sun hmdg-db1 [keepalived_notify]: umount /oradata success. Apr 15 17:35:05 Sun hmdg-db1 [keepalived_notify]: stopping listener... Apr 15 17:35:08 Sun hmdg-db1 [keepalived_notify]: The listener stop successfully
观察执行脚本的输出结果
1、 脚本接收转换到BACKUP信息
2、 首先移除VIP(由keepalived绑定到主机上浮动IP地址，同时也是对外提供服务的IP地址)，实际上是由keepalived检测到异常或本身的服务器不可用时自动剔除VIP
3、 备份控制文件到/backup/oracle/control/目录下
4、 正常关闭数据库实例
5、 卸载共享存储
6、 关闭数据库监听程序
l notify脚本切换演示
如果备机主动切换到MASTER状态时，脚本首先或检测对方磁盘是否umount，如果没有则最多等待20秒，提示异常并正常推出切换到MASTER的请求
[root@hmdg-db1 scripts]# ./keepalived_notify.sh master ------------------------------------------------------------------------------- Apr 15 17:40:08 Sun hmdg-db1 [keepalived_notify]: Transition to MASTER STATE Apr 15 17:40:08 Sun hmdg-db1 [keepalived_notify]: Setup the VIP on eth2 172.16.10.130,172.16.10.131 Apr 15 17:40:08 Sun hmdg-db1 [keepalived_notify]: Database Switchover To MASTER Apr 15 17:40:08 Sun hmdg-db1 [keepalived_notify]: Check remote node sharedisk mounted. Apr 15 17:40:08 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [1]... Apr 15 17:40:09 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [2]... Apr 15 17:40:10 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [3]... Apr 15 17:40:12 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [4]... Apr 15 17:40:13 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [5]... Apr 15 17:40:14 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [6]... Apr 15 17:40:15 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [7]... Apr 15 17:40:16 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [8]... Apr 15 17:40:17 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [9]... Apr 15 17:40:18 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [10]... Apr 15 17:40:20 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [11]... Apr 15 17:40:21 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [12]... Apr 15 17:40:22 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [13]... Apr 15 17:40:23 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [14]... Apr 15 17:40:24 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [15]... Apr 15 17:40:25 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [16]... Apr 15 17:40:26 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [17]... Apr 15 17:40:27 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [18]... Apr 15 17:40:29 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [19]... Apr 15 17:40:30 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [20]... Apr 15 17:40:30 Sun hmdg-db1 [keepalived_notify]: Disk status abnormal.
这时如果需要手动切换，则需要先将远程备用节点的服务关闭并umount存储之后，才能正常切换
[root@hmdg-db1 scripts]# ./keepalived_notify.sh master
-------------------------------------------------------------------------------
Apr 15 17:45:29 Sun hmdg-db1 [keepalived_notify]: Transition to MASTER STATE
Apr 15 17:45:29 Sun hmdg-db1 [keepalived_notify]: Setup the VIP on eth2 172.16.10.130,172.16.10.131
Apr 15 17:45:29 Sun hmdg-db1 [keepalived_notify]: Database Switchover To MASTER
Apr 15 17:45:29 Sun hmdg-db1 [keepalived_notify]: Check remote node sharedisk mounted.
Apr 15 17:45:29 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [1]...
Apr 15 17:45:30 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [2]...
Apr 15 17:45:31 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [3]...
Apr 15 17:45:32 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [4]...
Apr 15 17:45:33 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [5]...
Apr 15 17:45:35 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [6]...
Apr 15 17:45:36 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [7]...
Apr 15 17:45:37 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [8]...
Apr 15 17:45:38 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [9]...
Apr 15 17:45:39 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [10]...
Apr 15 17:45:40 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [11]...
Apr 15 17:45:41 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [12]...
Apr 15 17:45:42 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 is already mounted on remote node or busy. checking [13]...
Apr 15 17:45:44 Sun hmdg-db1 [keepalived_notify]: /dev/sdb1 check passed.
Apr 15 17:45:44 Sun hmdg-db1 [keepalived_notify]: mount /dev/sdb1 on /oradata
Apr 15 17:45:44 Sun hmdg-
网页名称：OracleHA双机主备基于共享存储模式并利用keepalived管理实现高可用
文章路径：http://kswsj.cn/article/jegsse.html
