0
0
1
0
博客/.../

基于 TiCDC 构建主备集群

 YangChao  发表于  2026-03-09
原创部署与初始化

服务器环境：

共享存储：192.168.68.100 (NFS、MinIO)
主用集群：192.168.68.101
备用集群：192.168.68.102

部署NFS

#服务器端 192.168.68.100
yum -y install nfs-utils
systemctl enable rpcbind.service
systemctl enable nfs-server.service
systemctl start rpcbind.service
systemctl start nfs-server
mkdir /data/nfsdata
vi /etc/exports
/data/nfsdata        192.168.68.0/24(rw,sync,no_root_squash) 

#客户端 192.168.68.101,192.168.68.102
yum -y install nfs-utils
showmount -e 192.168.68.100
mkdir /data/nfsdata
mount -t nfs 192.168.68.100:/data/nfsdata /data/nfsdata
chown -R tidb.tidb /data/nfsdata

#/etc/fstab
192.168.68.100:/data/nfsdata             /data/nfsdata                nfs     defaults       0 0

部署MinIO

#服务器端
https://dl.minio.org.cn/server/minio/release/linux-amd64/
#客户端
https://dl.minio.org.cn/client/mc/release/linux-amd64/

#官网：https://www.minio.org.cn/
wget -O /usr/local/bin/minio https://dl.minio.org.cn/server/minio/release/linux-amd64/minio # 服务端
wget -O /usr/local/bin/mc https://dl.minio.org.cn/client/mc/release/linux-amd64/mc # 客户端
chmod 755 /usr/local/bin/{minio,mc}
[root@localhost ~]# minio --version
[root@localhost ~]# mc --version

创建数据目录、日志目录，添加用户：

mkdir -p /data/minio/{data,logs,backup,redo}
useradd minio && echo minio | passwd --stdin minio # 可选，创建普通用户来运行进程
chown minio.minio -R /data/minio/

创建启动脚本

su - minio
vim /data/minio/minio.sh

#!/bin/bash
export MINIO_ROOT_USER=admin
export MINIO_ROOT_PASSWORD=Admin@123

MINIO_DATA_DIR=/data/minio/data
MINIO_LOG_DIR=/data/minio/logs
MINIO_ADDR=0.0.0.0 #监听端口和地址
MINIO_API=9000 #数据服务端口，用于处理用户的上传、下载等数据操作请求
MINIO_UI=9001  #控制台服务端口，用于提供Web管理界面

case "$1" in
  start)
      nohup minio server ${MINIO_DATA_DIR} \
            --console-address ${MINIO_ADDR}:${MINIO_UI} \
            --address ${MINIO_ADDR}:${MINIO_API} >> ${MINIO_LOG_DIR}/minio-$(date +%Y%m%d).log 2>&1 &
      ;;
  stop)
      pid=$(ps -ef | egrep "minio[/]" | awk '{print $2}')
      [ ! -z ${pid} ] && kill -9 ${pid} || echo "MinIO not running..."
      ;;    
  *)
      echo "Usage: $0 {start|stop}"
      exit 1
esac

启动服务

bash minio.sh start

访问页面:

http://192.168.68.100:9001/login (admin/Admin@123)

创建bucket:

在管理页面点击：Create Bucket 之后，输入名称创建：backup、redo

​image.png​

TiDB使用MinIO备份恢复示例

#备份
mysql> BACKUP DATABASE * TO 's3://backup/202507071609?access-key=admin&secret-access-key=Admin@123&endpoint=http://192.168.68.100:9000&force-path-style=true';
mysql> show backups;

#恢复
mysql> RESTORE DATABASE * FROM 's3://backup/202507071609?access-key=admin&secret-access-key=Admin@123&endpoint=http://192.168.68.100:9000&force-path-style=true';
mysql> show restores;

#连接MinIO
mc alias set myminio  http://192.168.68.100:9000 admin Admin@123
mc ls myminio
mc admin info myminio   

# mc mb myminio/mybucket02                 # 创建bucket
# mc ls myminio                            # 列出bucket
# mc rb myminio/mybucket02                 # 删除没有object的bucket
# mc rb myminio/mybucket02 --force         # 强制删除bucket，即使含有文件

#级联删除目录
mc rm myminio/tidbbackup/bak_20210922/ --recursive --force

#递归强制删除8天之前的文件，单位支持天，小时，分钟等
mc rm myminio/tidbbackup/fullbackup/ --recursive --force --older-than 8d

#操作系统命令备份
export AWS_ACCESS_KEY_ID=admin
export AWS_SECRET_ACCESS_KEY=Admin@123
tiup br backup table --db * --table * --ratelimit 60 --pd "192.168.68.101:2379" --storage "s3://backup/202507071635" --send-credentials-to-tikv=true --s3.endpoint "http://192.168.68.100:9000" --log-file ./backup.log

备份和恢复

备份表 db23.t1，限速 60M

export AWS_ACCESS_KEY_ID=myminioid
export AWS_SECRET_ACCESS_KEY=myminioPassWord
/data/br_513/br backup table --db db23 --table t1 --ratelimit 60 --pd "192.168.1.25:2379" --storage "s3://tidbbackup/bak_20220105" --send-credentials-to-tikv=true --s3.endpoint "http://192.168.1.32:9000" --log-file /tmp/backup.log

恢复表 db23.t1，限速 60M

export AWS_ACCESS_KEY_ID=myminioid
export AWS_SECRET_ACCESS_KEY=myminioPassWord
/data/br_513/br restore table --db db23 --table t1 --ratelimit 60 --pd "192.168.2.25:2379" --storage "s3://tidbbackup/bak_20220105" --send-credentials-to-tikv=true --s3.endpoint "http://192.168.1.32:9000" --log-file restore.log

部署tidb主备集群

分别在101和102上独立部署tidb集群(pd,tidb,tikv,tiflash,ticdc,monitor,grafana,altermanager)。

topology.yaml (192.168.68.101:tidb851a)

tiup cluster deploy tidb851a v8.5.1 topology.yaml --user tidb -p

# # Global variables are applied to all deployments and used as the default value of
# # the deployments if a specific deployment value is missing.
global:
 user: "tidb"
 ssh_port: 22
 deploy_dir: "/opt/tidb-deploy"
 data_dir: "/data/tidb-data"

# # Monitored variables are applied to all the machines.
monitored:
 node_exporter_port: 9100
 blackbox_exporter_port: 9115
 deploy_dir: "/opt/tidb-deploy/monitored-9100"
 data_dir: "/data/tidb-data/monitored-9100"
 log_dir: "/opt/tidb-deploy/monitored-9100/log"

server_configs:
 tidb:
   log.slow-threshold: 300
 tikv:
   readpool.storage.use-unified-pool: false
   readpool.coprocessor.use-unified-pool: true
 pd:
   replication.enable-placement-rules: true
   replication.location-labels: ["zone","idc","host"]
   replication.isolation-level: "host"
 tiflash:
   logger.level: "info"

pd_servers:
 - host: 192.168.68.101
   name: "pd1"
   client_port: 2379
   peer_port: 2380
   deploy_dir: "/opt/tidb-deploy/pd-2379"
   data_dir: "/data/tidb-data/pd-2379"
   log_dir: "/opt/tidb-deploy/pd-2379/log"

tidb_servers:
 - host: 192.168.68.101
   port: 4000
   status_port: 10080
   deploy_dir: "/opt/tidb-deploy/tidb-4000"
   log_dir: "/opt/tidb-deploy/tidb-4000/log"

tikv_servers:
 - host: 192.168.68.101
   port: 20161
   status_port: 20181
   deploy_dir: "/opt/tidb-deploy/tikv-20161"
   data_dir: "/data1/tidb-data/tikv-20161"
   log_dir: "/opt/tidb-deploy/tikv-20161/log"
   config:
     server.labels: { zone: "zone1", idc: "idc1", host: "host1" }

 - host: 192.168.68.101
   port: 20162
   status_port: 20182
   deploy_dir: "/opt/tidb-deploy/tikv-20162"
   data_dir: "/data2/tidb-data/tikv-20162"
   log_dir: "/opt/tidb-deploy/tikv-20162/log"
   config:
     server.labels: { zone: "zone1", idc: "idc1", host: "host2" }

 - host: 192.168.68.101
   port: 20163
   status_port: 20183
   deploy_dir: "/opt/tidb-deploy/tikv-20163"
   data_dir: "/data3/tidb-data/tikv-20163"
   log_dir: "/opt/tidb-deploy/tikv-20163/log"
   config:
     server.labels: { zone: "zone1", idc: "idc1", host: "host3" }

tiflash_servers:
 - host: 192.168.68.101
   tcp_port: 9000
   flash_service_port: 3930
   flash_proxy_port: 20170
   flash_proxy_status_port: 20292
   metrics_port: 8234
   deploy_dir: "/opt/tidb-deploy/tiflash-9000"
   data_dir: "/data/tidb-data/tiflash-9000"
   log_dir: "/opt/tidb-deploy/tiflash-9000/log"

monitoring_servers:
 - host: 192.168.68.101
   port: 9090
   deploy_dir: "/opt/tidb-deploy/prometheus-8249"
   data_dir: "/data/tidb-data/prometheus-8249"
   log_dir: "/opt/tidb-deploy/prometheus-8249/log"

grafana_servers:
 - host: 192.168.68.101
   port: 3000
   deploy_dir: /opt/tidb-deploy/grafana-3000

alertmanager_servers:
  - host: 192.168.68.101
    web_port: 9093
    cluster_port: 9094
    deploy_dir: "/opt/tidb-deploy/alertmanager-9093"
    data_dir: "/data/tidb-data/alertmanager-9093"
    log_dir: "/opt/tidb-deploy/alertmanager-9093/log"

cdc_servers:
  - host: 192.168.68.101
    port: 8300
    gc-ttl: 86400
    deploy_dir: "/opt/tidb-deploy/cdc-8300"
    data_dir: "/data/tidb-data/cdc-8300"
    log_dir: "/opt/tidb-deploy/cdc-8300/log"

topology.yaml (192.168.68.102:tidb851b)

tiup cluster deploy tidb851b v8.5.1 topology.yaml --user tidb -p

# # Global variables are applied to all deployments and used as the default value of
# # the deployments if a specific deployment value is missing.
global:
  user: "tidb"
  ssh_port: 22
  deploy_dir: "/opt/tidb-deploy"
  data_dir: "/data/tidb-data"

# # Monitored variables are applied to all the machines.
monitored:
  node_exporter_port: 9100
  blackbox_exporter_port: 9115
  deploy_dir: "/opt/tidb-deploy/monitored-9100"
  data_dir: "/data/tidb-data/monitored-9100"
  log_dir: "/opt/tidb-deploy/monitored-9100/log"

server_configs:
  tidb:
    log.slow-threshold: 300
  tikv:
    readpool.storage.use-unified-pool: false
    readpool.coprocessor.use-unified-pool: true
  pd:
    replication.enable-placement-rules: true
    replication.location-labels: ["zone","idc","host"]
    replication.isolation-level: "host"
  tiflash:
    logger.level: "info"

pd_servers:
  - host: 192.168.68.102
    name: "pd1"
    client_port: 2379
    peer_port: 2380
    deploy_dir: "/opt/tidb-deploy/pd-2379"
    data_dir: "/data/tidb-data/pd-2379"
    log_dir: "/opt/tidb-deploy/pd-2379/log"

tidb_servers:
  - host: 192.168.68.102
    port: 4000
    status_port: 10080
    deploy_dir: "/opt/tidb-deploy/tidb-4000"
    log_dir: "/opt/tidb-deploy/tidb-4000/log"

tikv_servers:
  - host: 192.168.68.102
    port: 20161
    status_port: 20181
    deploy_dir: "/opt/tidb-deploy/tikv-20161"
    data_dir: "/data1/tidb-data/tikv-20161"
    log_dir: "/opt/tidb-deploy/tikv-20161/log"
    config:
      server.labels: { zone: "zone2", idc: "idc1", host: "host1" }

  - host: 192.168.68.102
    port: 20162
    status_port: 20182
    deploy_dir: "/opt/tidb-deploy/tikv-20162"
    data_dir: "/data2/tidb-data/tikv-20162"
    log_dir: "/opt/tidb-deploy/tikv-20162/log"
    config:
      server.labels: { zone: "zone2", idc: "idc1", host: "host2" }

  - host: 192.168.68.102
    port: 20163
    status_port: 20183
    deploy_dir: "/opt/tidb-deploy/tikv-20163"
    data_dir: "/data3/tidb-data/tikv-20163"
    log_dir: "/opt/tidb-deploy/tikv-20163/log"
    config:
      server.labels: { zone: "zone2", idc: "idc1", host: "host3" }

tiflash_servers:
  - host: 192.168.68.102
    tcp_port: 9000
    flash_service_port: 3930
    flash_proxy_port: 20170
    flash_proxy_status_port: 20292
    metrics_port: 8234
    deploy_dir: "/opt/tidb-deploy/tiflash-9000"
    data_dir: "/data/tidb-data/tiflash-9000"
    log_dir: "/opt/tidb-deploy/tiflash-9000/log"

monitoring_servers:
  - host: 192.168.68.102
    port: 9090
    deploy_dir: "/opt/tidb-deploy/prometheus-8249"
    data_dir: "/data/tidb-data/prometheus-8249"
    log_dir: "/opt/tidb-deploy/prometheus-8249/log"

grafana_servers:
  - host: 192.168.68.102
    port: 3000
    deploy_dir: /opt/tidb-deploy/grafana-3000

alertmanager_servers:
  - host: 192.168.68.102
    web_port: 9093
    cluster_port: 9094
    deploy_dir: "/opt/tidb-deploy/alertmanager-9093"
    data_dir: "/data/tidb-data/alertmanager-9093"
    log_dir: "/opt/tidb-deploy/alertmanager-9093/log"

cdc_servers:
  - host: 192.168.68.102
    port: 8300
    gc-ttl: 86400
    deploy_dir: "/opt/tidb-deploy/cdc-8300"
    data_dir: "/data/tidb-data/cdc-8300"
    log_dir: "/opt/tidb-deploy/cdc-8300/log"

迁移全量数据

  1. 关闭 GC。为了保证增量迁移过程中新写入的数据不丢失，在开始备份之前，需要关闭上游集群的垃圾回收 (GC) 机制，以确保系统不再清理历史数据。执行如下命令关闭 GC：

    mysql> SET GLOBAL tidb_gc_enable=FALSE;
Query OK, 0 rows affected (0.01 sec)

    查询 tidb_gc_enable​ 的取值，判断 GC 是否已关闭：

    mysql> SELECT @@global.tidb_gc_enable;
+-------------------------+
| @@global.tidb_gc_enable |
+-------------------------+
|                       0 |
+-------------------------+
1 row in set (0.00 sec)

  2. 备份数据。在上游集群中执行 BACKUP 语句备份数据(NFS)：

    mysql> backup database * to 'local:///data/nfsdata/tidb851a-backup/';
+----------------------------------------+--------+--------------------+---------------------+---------------------+
| Destination                            | Size   | BackupTS           | Queue Time          | Execution Time      |
+----------------------------------------+--------+--------------------+---------------------+---------------------+
| local:///data/nfsdata/tidb851a-backup/ | 118560 | 459081695216795655 | 2025-06-30 12:28:07 | 2025-06-30 12:28:07 |
+----------------------------------------+--------+--------------------+---------------------+---------------------+
1 row in set (3.95 sec)

  3. 备份语句提交成功后，TiDB 会返回关于备份数据的元信息，这里需要重点关注 BackupTS，它意味着该时间点之前数据会被备份，后边的教程中，本文将使用 BackupTS 作为数据校验截止时间和 TiCDC 增量扫描的开始时间。

  4. 恢复数据。在下游集群中执行 RESTORE 语句恢复数据：

    mysql> use testdb;
Database changed
mysql> show tables;
Empty set (0.00 sec)

mysql> restore database * from 'local:///data/nfsdata/tidb851a-backup/';
+----------------------------------------+--------+--------------------+--------------------+---------------------+---------------------+
| Destination                            | Size   | BackupTS           | Cluster TS         | Queue Time          | Execution Time      |
+----------------------------------------+--------+--------------------+--------------------+---------------------+---------------------+
| local:///data/nfsdata/tidb851a-backup/ | 118560 | 459081695216795655 | 459081711796879364 | 2025-06-30 12:29:10 | 2025-06-30 12:29:10 |
+----------------------------------------+--------+--------------------+--------------------+---------------------+---------------------+
1 row in set (2.82 sec)

mysql> show tables;
+------------------+
| Tables_in_testdb |
+------------------+
| t1               |
| t2               |
| t3               |
| t4               |
+------------------+
4 rows in set (0.00 sec)

  5. （可选）校验数据。通过 sync-diff-inspector 工具，可以验证上下游数据在某个时间点的一致性。从上述备份和恢复命令的输出可以看到，上游集群备份的时间点为 431434047157698561，下游集群完成数据恢复的时间点为 431434141450371074。

    sync_diff_inspector -C ./config.yaml

    关于 sync-diff-inspector 的配置方法，请参考配置文件说明。在本文中，相应的配置如下：

    # Diff Configuration.
######################### Global config #########################
check-thread-count = 4
export-fix-sql = true
check-struct-only = false

######################### Datasource config #########################
[data-sources]
[data-sources.upstream]
    host = "192.168.68.101" # 替换为实际上游集群 ip
    port = 4000
    user = "root"
    password = "root"
    snapshot = "459081695216795655" # 配置为实际的备份时间点
[data-sources.downstream]
    host = "192.168.68.102" # 替换为实际下游集群 ip
    port = 4000
    user = "root"
    password = "root"
    snapshot = "459081695216795655" # 配置为实际的恢复时间点

######################### Task config #########################
[task]
    output-dir = "./output"
    source-instances = ["upstream"]
    target-instance = "downstream"
    target-check-tables = ["*.*"]

迁移增量数据

  1. 部署 TiCDC。完成全量数据迁移后，就可以部署并配置 TiCDC 集群同步增量数据，实际生产集群中请参考 TiCDC 部署。本文在创建测试集群时，已经启动了一个 TiCDC 节点，因此可以直接进行 changefeed 的配置。

  2. 创建同步任务。创建 changefeed 配置文件并保存为 changefeed.toml​。

    [consistent]
# 一致性级别，配置成 eventual 表示开启一致性复制
level = "eventual"
# 使用 S3 来存储 redo log, 其他可选为 local, nfs
storage = "local:///data/nfsdata/tidb851a-redo"

    在上游集群中(192.168.68.101)，执行以下命令创建从上游到下游集群的同步链路：

    tiup cdc cli changefeed create --server=http://192.168.68.101:8300 \
--sink-uri="mysql://root:root@192.168.68.102:4000" \
--changefeed-id="tidb851a-to-tidb851b" \
--start-ts="459081695216795655" \
--config=changefeed.toml

    以上命令中：

    • ​--server​：TiCDC 集群任意一节点的地址
    • ​--sink-uri​：同步任务下游的地址
    • ​--start-ts​：TiCDC 同步的起点，需要设置为实际的备份时间点（也就是第 2 步：迁移全量数据提到的 BackupTS）

  3. 更多关于 changefeed 的配置，请参考 TiCDC Changefeed 配置参数。

  4. 重新开启 GC。TiCDC 可以保证未同步的历史数据不会被回收。因此，创建完从上游到下游集群的 changefeed 之后，就可以执行如下命令恢复集群的垃圾回收功能。详情请参考 TiCDC GC safepoint 的完整行为。执行如下命令打开 GC：

    mysql> SET GLOBAL tidb_gc_enable=TRUE;
Query OK, 0 rows affected (0.03 sec)

    查询 tidb_gc_enable​ 的取值，判断 GC 是否已开启：

    mysql> SELECT @@global.tidb_gc_enable;
+-------------------------+
| @@global.tidb_gc_enable |
+-------------------------+
|                       1 |
+-------------------------+
1 row in set (0.00 sec)

验证数据同步

在主用集群上执行DM后，查询备用集群数据是否同步。

#主
mysql> use testdb;
mysql> insert into t4 values(2,'testcdc');

#备
mysql> use testdb;
mysql> select * from t4;

同步任务操作任务参考：

#查看同步任务

tiup cdc cli changefeed list --server=http://192.168.68.101:8300

#查询特定同步任务

tiup cdc cli changefeed query --server=http://192.168.68.101:8300 --changefeed-id=tidb851a-to-tidb851b
tiup cdc cli changefeed query -s --server=http://192.168.68.101:8300 --changefeed-id=tidb851a-to-tidb851b

#停止同步任务

tiup cdc cli changefeed pause --server=http://192.168.68.101:8300 --changefeed-id=tidb851a-to-tidb851b

#恢复同步任务

tiup cdc cli changefeed resume --server=http://192.168.68.101:8300 --changefeed-id tidb851a-to-tidb851b

#删除同步任务

tiup cdc cli changefeed remove --server=http://192.168.68.101:8300 --changefeed-id tidb851a-to-tidb851b

MinIO常用相关命令

#连接minio
mc alias set myminio http://192.168.68.100:9000 admin Admin@123
mc admin info myminio

#创建存储桶
mc mb backup/tidb851a20250708

#查看存储桶
mc ls myminio

#删除没有文件的bucket=>  mc rb  链接名/存储桶
mc rb test/test

# 删除有文件的bucket => mc rb  链接名/存储桶 --force
mc rb test/test --force

0
0
1
0

版权声明：本文为 TiDB 社区用户原创文章，遵循 CC BY-NC-SA 4.0 版权协议，转载请附上原文出处链接和本声明。

评论
暂无评论
备份和恢复