# 步骤1：部署OGG集群
# 节点1配置
$ mkdir -p /GoldenGate/app
$ chown ogg:ogg /GoldenGate/app
$ cd /GoldenGate/app
$ unzip /software/fbo_ggs_Linux_x64_shiphome.zip
$ ./ggsci

GGSCI (ogg-node1)> CREATE SUBDIRS

# 节点2配置（相同步骤）
$ mkdir -p /GoldenGate/app
$ chown ogg:ogg /GoldenGate/app
$ cd /GoldenGate/app
$ unzip /software/fbo_ggs_Linux_x64_shiphome.zip
$ ./ggsci

GGSCI (ogg-node2)> CREATE SUBDIRS

# 步骤2：配置共享存储
# 在两个节点挂载共享存储
$ mkdir -p /GoldenGate/shared
$ mount -t nfs nas-server:/ogg-shared /GoldenGate/shared

# 配置Trail文件存储到共享存储
GGSCI (ogg-node1)> EDIT PARAMS E_FGEDU01

EXTRACT E_FGEDU01
USERIDALIAS fgedu_alias
EXTTRAIL /GoldenGate/shared/dirdat/ea
TABLE fgedu01.*;

# 步骤3：配置Manager高可用
GGSCI (ogg-node1)> EDIT PARAMS MGR

PORT 7809
PURGEOLDEXTRACTS /GoldenGate/shared/dirdat/*, USECHECKPOINTS

— 自动重启配置
AUTORESTART EXTRACT *, RETRIES 5, WAITMINUTES 2, RESETMINUTES 60
AUTORESTART REPLICAT *, RETRIES 5, WAITMINUTES 2, RESETMINUTES 60

— 故障转移配置
FAILOVER DYNAMIC

# 步骤4：配置检查点表
SQL> CREATE USER fgedu IDENTIFIED BY fgedu123;
SQL> GRANT CONNECT, RESOURCE, DBA TO fgedu;

GGSCI (ogg-node1)> DBLOGIN USERIDALIAS fgedu_alias
GGSCI (ogg-node1)> ADD CHECKPOINTTABLE fgedu.ogg_chkpt

# 步骤5：配置多路径Trail
GGSCI (ogg-node1)> EDIT PARAMS E_FGEDU01

EXTRACT E_FGEDU01
USERIDALIAS fgedu_alias
EXTTRAIL /GoldenGate/shared/dirdat/ea
ALTTRAIL /GoldenGate/local/dirdat/ea
TABLE fgedu01.*;

# 步骤6：配置Data Pump冗余
GGSCI (ogg-node1)> EDIT PARAMS P_FGEDU01

EXTRACT P_FGEDU01
USERIDALIAS fgedu_alias
RMTHOST ogg-node2, MGRPORT 7809
ALTHOST ogg-node3, ALTPORT 7809
RMTTRAIL /GoldenGate/shared/dirdat/ra
TABLE fgedu01.*;

# 步骤7：启动OGG集群
# 节点1
GGSCI (ogg-node1)> START MANAGER
GGSCI (ogg-node1)> START EXTRACT E_FGEDU01
GGSCI (ogg-node1)> START EXTRACT P_FGEDU01

# 节点2
GGSCI (ogg-node2)> START MANAGER
GGSCI (ogg-node2)> START REPLICAT R_FGEDU01

# 步骤8：验证集群状态
GGSCI (ogg-node1)> INFO ALL

Program Status Group Lag at Chkpt Time Since Chkpt

MANAGER RUNNING
EXTRACT RUNNING E_FGEDU01 00:00:00 00:00:05
EXTRACT RUNNING P_FGEDU01 00:00:00 00:00:03

# 步骤1：配置Manager自动重启
GGSCI (fgedu.net.cn)> EDIT PARAMS MGR

PORT 7809
PURGEOLDEXTRACTS /GoldenGate/app/dirdat/*, USECHECKPOINTS

— Extract进程自动重启
AUTORESTART EXTRACT E_*, RETRIES 5, WAITMINUTES 2, RESETMINUTES 60
AUTORESTART EXTRACT P_*, RETRIES 5, WAITMINUTES 2, RESETMINUTES 60

— Replicat进程自动重启
AUTORESTART REPLICAT R_*, RETRIES 5, WAITMINUTES 2, RESETMINUTES 60

— 启动延迟配置
STARTUPVALIDATIONDELAY 30

# 步骤2：创建进程监控脚本
#!/bin/bash
# ogg_ha_monitor.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

OGG_HOME=/GoldenGate/app
LOG_FILE=/GoldenGate/app/logs/ha_monitor.log
MAX_RETRIES=3

log_message() {
echo “[$(date ‘+%Y-%m-%d %H:%M:%S’)] $1” >> $LOG_FILE
}

# 检查进程状态
check_process() {
local process_name=$1
cd $OGG_HOME

status=$(echo “INFO EXTRACT $process_name 2>/dev/null || INFO REPLICAT $process_name” | ./ggsci | grep “Status” | awk ‘{print $2}’)

if [ “$status” != “RUNNING” ]; then
return 1
fi
return 0
}

# 重启进程
restart_process() {
local process_name=$1
cd $OGG_HOME

# 判断进程类型
if echo “INFO ALL” | ./ggsci | grep “EXTRACT.*$process_name” > /dev/null; then
echo “START EXTRACT $process_name” | ./ggsci
else
echo “START REPLICAT $process_name” | ./ggsci
fi
}

# 主监控循环
main() {
log_message “开始OGG高可用监控”

cd $OGG_HOME
processes=$(echo “INFO ALL” | ./ggsci | grep -E “EXTRACT|REPLICAT” | awk ‘{print $3}’)

for proc in $processes; do
if ! check_process $proc; then
log_message “进程 $proc 异常，尝试重启”

retry=0
while [ $retry -lt $MAX_RETRIES ]; do
retry=$((retry + 1))
log_message “第 $retry 次尝试重启 $proc”

restart_process $proc
sleep 10

if check_process $proc; then
log_message “进程 $proc 重启成功”
break
fi
done

if [ $retry -eq $MAX_RETRIES ]; then
log_message “进程 $proc 重启失败，发送告警”
# 发送告警
/GoldenGate/app/scripts/alert/send_alert.sh “P1” “OGG进程重启失败” “$proc 重启失败，需要人工介入”
fi
fi
done

log_message “OGG高可用监控完成”
}

main

# 步骤3：设置定时任务
$ crontab -e
*/1 * * * * /GoldenGate/app/scripts/ogg_ha_monitor.sh >> /GoldenGate/app/logs/cron.log 2>&1

# 步骤1：数据中心A配置
GGSCI (dc-a.fgedu.net.cn)> EDIT PARAMS E_TO_B

EXTRACT E_TO_B
USERIDALIAS fgedu_alias
EXTTRAIL /GoldenGate/app/dirdat/ab
TABLE fgedu01.*;
TABLE fgedu02.*;

GGSCI (dc-a.fgedu.net.cn)> EDIT PARAMS R_FROM_B

REPLICAT R_FROM_B
USERIDALIAS fgedu_alias
ASSUMETARGETDEFS
MAP fgedu01.*, TARGET fgedu01.*;
MAP fgedu02.*, TARGET fgedu02.*;

— 避免循环复制
TRANLOGOPTIONS EXCLUDETAG 99

# 步骤2：数据中心B配置
GGSCI (dc-b.fgedu.net.cn)> EDIT PARAMS E_TO_A

EXTRACT E_TO_A
USERIDALIAS fgedu_alias
EXTTRAIL /GoldenGate/app/dirdat/ba
TABLE fgedu01.*;
TABLE fgedu02.*;

GGSCI (dc-b.fgedu.net.cn)> EDIT PARAMS R_FROM_A

REPLICAT R_FROM_A
USERIDALIAS fgedu_alias
ASSUMETARGETDEFS
MAP fgedu01.*, TARGET fgedu01.*;
MAP fgedu02.*, TARGET fgedu02.*;

— 避免循环复制
TRANLOGOPTIONS EXCLUDETAG 99

# 步骤3：创建进程
# 数据中心A
GGSCI (dc-a)> ADD EXTRACT E_TO_B, TRANLOG, BEGIN NOW
GGSCI (dc-a)> ADD EXTTRAIL /GoldenGate/app/dirdat/ab, EXTRACT E_TO_B
GGSCI (dc-a)> ADD REPLICAT R_FROM_B, EXTTRAIL /GoldenGate/app/dirdat/ba, CHECKPOINTTABLE fgedu.ogg_chkpt

# 数据中心B
GGSCI (dc-b)> ADD EXTRACT E_TO_A, TRANLOG, BEGIN NOW
GGSCI (dc-b)> ADD EXTTRAIL /GoldenGate/app/dirdat/ba, EXTRACT E_TO_A
GGSCI (dc-b)> ADD REPLICAT R_FROM_A, EXTTRAIL /GoldenGate/app/dirdat/ab, CHECKPOINTTABLE fgedu.ogg_chkpt

# 步骤4：启动双向同步
# 数据中心A
GGSCI (dc-a)> START EXTRACT E_TO_B
GGSCI (dc-a)> START REPLICAT R_FROM_B

# 数据中心B
GGSCI (dc-b)> START EXTRACT E_TO_A
GGSCI (dc-b)> START REPLICAT R_FROM_A

# 步骤5：验证双向同步
# 在数据中心A插入数据
SQL> INSERT INTO fgedu01.fgedu_orders VALUES (1, ‘DC-A’, SYSDATE);
SQL> COMMIT;

# 在数据中心B验证
SQL> SELECT * FROM fgedu01.fgedu_orders WHERE id = 1;

ID SOURCE CREATE_TIME
———- ———- ——————-
1 DC-A 2026-04-10 10:00:00

# 在数据中心B插入数据
SQL> INSERT INTO fgedu01.fgedu_orders VALUES (2, ‘DC-B’, SYSDATE);
SQL> COMMIT;

# 在数据中心A验证
SQL> SELECT * FROM fgedu01.fgedu_orders WHERE id = 2;

ID SOURCE CREATE_TIME
———- ———- ——————-
2 DC-B 2026-04-10 10:00:05
from GoldenGate视频:www.itpux.com

# 步骤1：配置冲突检测
GGSCI (dc-a.fgedu.net.cn)> EDIT PARAMS R_FROM_B

REPLICAT R_FROM_B
USERIDALIAS fgedu_alias
ASSUMETARGETDEFS

— 冲突处理配置
MAP fgedu01.fgedu_orders, TARGET fgedu01.fgedu_orders,
COMPARECOLS (ALL, EXCEPT create_time),
RESOLVECONFLICT (UPDATEROWEXISTS, (DEFAULT, OVERWRITE)),
RESOLVECONFLICT (INSERTROWEXISTS, (DEFAULT, OVERWRITE)),
RESOLVECONFLICT (DELETEROWEXISTS, (DEFAULT, OVERWRITE)),
RESOLVECONFLICT (UPDATEROWMISSING, (DEFAULT, DISCARD)),
RESOLVECONFLICT (DELETEROWMISSING, (DEFAULT, DISCARD));

— 记录冲突日志
REPORTCOUNT EVERY 1 MINUTES, RATE
REPORTROLLOVER AT 00:00

# 步骤2：配置异常处理表
SQL> CREATE TABLE fgedu.ogg_conflicts (
conflict_time TIMESTAMP,
table_name VARCHAR2(100),
operation VARCHAR2(10),
conflict_type VARCHAR2(50),
source_data CLOB,
target_data CLOB,
resolution VARCHAR2(50)
);

# 步骤3：配置冲突处理函数
GGSCI (dc-a.fgedu.net.cn)> EDIT PARAMS R_FROM_B

REPLICAT R_FROM_B
USERIDALIAS fgedu_alias
ASSUMETARGETDEFS

— 使用存储过程处理冲突
MAP fgedu01.fgedu_orders, TARGET fgedu01.fgedu_orders,
COMPARECOLS (ALL, EXCEPT create_time),
RESOLVECONFLICT (UPDATEROWEXISTS, (DEFAULT, USEMAX (update_time)));

# 步骤4：创建冲突监控脚本
#!/bin/bash
# conflict_monitor.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

OGG_HOME=/GoldenGate/app
LOG_FILE=/GoldenGate/app/logs/conflict.log

check_conflicts() {
# 检查冲突日志
cd $OGG_HOME

for report in dirrpt/R_*.rpt; do
conflicts=$(grep -c “CONFLICT” $report 2>/dev/null)
if [ $conflicts -gt 0 ]; then
echo “[$(date)] 发现 $conflicts 个冲突: $report” >> $LOG_FILE

# 发送告警
/GoldenGate/app/scripts/alert/send_alert.sh “P2” “OGG数据冲突” “发现 $conflicts 个数据冲突”
fi
done
}

check_conflicts

# 步骤1：创建切换脚本
#!/bin/bash
# ogg_failover.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

ACTION=$1
LOG_FILE=/GoldenGate/app/logs/failover.log

log_message() {
echo “[$(date ‘+%Y-%m-%d %H:%M:%S’)] $1” >> $LOG_FILE
}

# 切换到备中心
failover_to_standby() {
log_message “开始切换到备中心”

# 1. 停止主中心OGG进程
log_message “停止主中心OGG进程”
echo “STOP EXTRACT *” | ggsci
echo “STOP REPLICAT *” | ggsci

# 2. 等待Trail文件传输完成
log_message “等待Trail文件传输完成”
sleep 30

# 3. 验证数据一致性
log_message “验证数据一致性”
# 执行数据校验脚本

# 4. 启动备中心OGG进程
log_message “启动备中心OGG进程”
ssh standby-dc “echo ‘START EXTRACT *’ | ggsci”
ssh standby-dc “echo ‘START REPLICAT *’ | ggsci”

# 5. 切换应用VIP
log_message “切换应用VIP”
# VIP切换脚本

# 6. 验证服务
log_message “验证服务状态”
# 服务验证脚本

log_message “切换完成”
}

# 切换回主中心
failback_to_primary() {
log_message “开始切换回主中心”

# 1. 确认主中心已恢复
log_message “确认主中心已恢复”

# 2. 启动反向同步
log_message “启动反向同步”
echo “START EXTRACT E_TO_PRIMARY” | ggsci

# 3. 等待数据追平
log_message “等待数据追平”
sleep 60

# 4. 验证数据一致性
log_message “验证数据一致性”

# 5. 停止备中心写入
log_message “停止备中心写入”
ssh standby-dc “echo ‘STOP EXTRACT *’ | ggsci”

# 6. 切换应用VIP
log_message “切换应用VIP”

# 7. 启动主中心OGG进程
log_message “启动主中心OGG进程”
echo “START EXTRACT *” | ggsci
echo “START REPLICAT *” | ggsci

log_message “切换完成”
}

case “$ACTION” in
failover)
failover_to_standby
;;
failback)
failback_to_primary
;;
*)
echo “用法: $0 {failover|failback}”
;;
esac

# 步骤2：执行切换演练
$ ./ogg_failover.sh failover

[2026-04-10 10:00:00] 开始切换到备中心
[2026-04-10 10:00:01] 停止主中心OGG进程
[2026-04-10 10:00:30] 等待Trail文件传输完成
[2026-04-10 10:01:00] 验证数据一致性
[2026-04-10 10:01:30] 启动备中心OGG进程
[2026-04-10 10:02:00] 切换应用VIP
[2026-04-10 10:02:30] 验证服务状态
[2026-04-10 10:03:00] 切换完成

# 步骤3：验证切换结果
# 检查备中心进程状态
GGSCI (standby-dc)> INFO ALL

Program Status Group Lag at Chkpt Time Since Chkpt

MANAGER RUNNING
EXTRACT RUNNING E_FGEDU01 00:00:00 00:00:05
REPLICAT RUNNING R_FGEDU01 00:00:00 00:00:03

# 检查数据同步状态
SQL> SELECT COUNT(*) FROM fgedu01.fgedu_orders;

COUNT(*)
———-
10000

# 项目背景
– 业务：银行核心交易系统
– 要求：RTO<5分钟，RPO=0 - 数据量：约10TB - 交易量：约10000 TPS # 架构设计 两地三中心架构： 北京主中心（生产） 北京备中心（双活） 上海灾备中心 +------------------+ +------------------+ +------------------+ | Oracle RAC | | Oracle RAC | | Oracle单机 | | 4节点 | | 4节点 | | 2节点 | +------------------+ +------------------+ +------------------+ | | | +------------------+ +------------------+ +------------------+ | OGG集群 |<---->| OGG集群 |—–>| OGG单机 |
| 双向同步 | | 双向同步 | | 单向同步 |
+——————+ +——————+ +——————+

# 关键配置
# 1. 双向同步配置
— 北京主中心
EXTRACT E_TO_BJ_B
EXTTRAIL /GoldenGate/app/dirdat/bb
TABLE FGCORE.*;

REPLICAT R_FROM_BJ_B
ASSUMETARGETDEFS
TRANLOGOPTIONS EXCLUDETAG 99
MAP FGCORE.*, TARGET FGCORE.*;

# 2. 异步复制到灾备
EXTRACT P_TO_SH
RMTHOST sh-dr-center, MGRPORT 7809
RMTTRAIL /GoldenGate/app/dirdat/sh
TABLE FGCORE.*;

# 3. 冲突处理
MAP FGCORE.TRANS_LOG, TARGET FGCORE.TRANS_LOG,
RESOLVECONFLICT (UPDATEROWEXISTS, (DEFAULT, USEMAX (TRANS_TIME)));

# 实施效果
| 指标 | 实施前 | 实施后 |
|—————|————-|————-|
| RTO | 4小时 | 3分钟 |
| RPO | 1小时 | 0 |
| 可用性 | 99.9% | 99.999% |
| 年故障时间 | 8.76小时 | 5.26分钟 |

# 运维要点
1. 每日检查同步状态
2. 每周进行切换演练
3. 每月进行灾备演练
4. 每季度进行架构评审

# 问题1：脑裂问题
# 现象
两个数据中心同时认为自己是主中心

# 原因分析
网络分区导致通信中断

# 解决方案
# 1. 配置仲裁机制
— 使用第三方仲裁服务器
FAILOVER DYNAMIC RESOLVE MASTER ogg-arbiter, PORT 7809

# 2. 配置心跳检测
— 配置多个心跳链路
RMTHOST host1, MGRPORT 7809, ALTHOST host2, ALTPORT 7809

# 问题2：数据冲突
# 现象
双向同步时出现数据不一致

# 原因分析
同一记录在两个中心同时修改

# 解决方案
# 1. 配置冲突处理规则
RESOLVECONFLICT (UPDATEROWEXISTS, (DEFAULT, USEMAX (UPDATE_TIME)))

# 2. 业务层面避免
– 按用户ID分片
– 按业务类型分片
– 使用分布式锁

# 问题3：切换后数据丢失
# 现象
切换后发现部分数据丢失

# 原因分析
Trail文件未完全传输

# 解决方案
# 1. 切换前检查同步状态
GGSCI> LAG REPLICAT R_FGEDU01

# 2. 使用检查点表
CHECKPOINTTABLE fgedu.ogg_chkpt

# 3. 验证数据一致性后再切换

# 解决方案1：完整高可用监控脚本
#!/bin/bash
# ogg_ha_complete.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

OGG_HOME=/GoldenGate/app
LOG_FILE=/GoldenGate/app/logs/ha_complete.log

log_message() {
echo “[$(date ‘+%Y-%m-%d %H:%M:%S’)] $1” >> $LOG_FILE
}

# 健康检查
health_check() {
log_message “开始健康检查”

# 检查进程状态
processes=$(echo “INFO ALL” | ggsci | grep -E “EXTRACT|REPLICAT” | awk ‘{print $3}’)

for proc in $processes; do
status=$(echo “INFO $proc” | ggsci | grep “Status” | awk ‘{print $2}’)

if [ “$status” != “RUNNING” ]; then
log_message “进程 $proc 状态异常: $status”
return 1
fi
done

# 检查延迟
for proc in $processes; do
lag=$(echo “LAG $proc” | ggsci | grep “Last record lag” | awk ‘{print $4}’)

if [ -n “$lag” ]; then
lag_seconds=$(echo $lag | awk -F: ‘{print ($1*3600)+($2*60)+$3}’)
if [ $lag_seconds -gt 300 ]; then
log_message “进程 $proc 延迟过大: $lag”
return 1
fi
fi
done

# 检查存储空间
disk_usage=$(df -h /GoldenGate | tail -1 | awk ‘{print $5}’ | cut -d% -f1)
if [ $disk_usage -gt 80 ]; then
log_message “存储空间不足: ${disk_usage}%”
return 1
fi

log_message “健康检查通过”
return 0
}

# 自动故障处理
auto_failover() {
log_message “开始自动故障处理”

# 检查是否需要切换
if ! health_check; then
log_message “检测到故障，准备切换”

# 通知备中心准备接管
ssh standby-dc “/GoldenGate/app/scripts/prepare_takeover.sh”

# 执行切换
/GoldenGate/app/scripts/ogg_failover.sh failover
fi
}

# 主程序
case “$1” in
check)
health_check
;;
failover)
auto_failover
;;
*)
echo “用法: $0 {check|failover}”
;;
esac

# 解决方案2：数据一致性验证脚本
#!/bin/bash
# data_consistency_check.sh
# from:www.itpux.com.qq113257174.wx:itpux-com
# web: http://www.fgedu.net.cn

PRIMARY_DB=”fgedu@primary-db”
STANDBY_DB=”fgedu@standby-db”
LOG_FILE=/GoldenGate/app/logs/consistency.log

check_table_count() {
local table=$1

primary_count=$(sqlplus -s $PRIMARY_DB <

Part05-风哥经验总结与分享

5.1 高可用最佳实践

根据多年高可用架构实施经验，总结以下最佳实践：