作者:沈光宇
爱可生南区 DBA 团队成员,主要负责 MySQL 故障处理和性能优化。对技术执着,为客户负责。
本文来源:原创投稿
*爱可生开源社区出品,原创内容未经授权不得随意使用,转载请联系小编并注明来源。
一、背景
某客户MySQL实例中数据较大(上T的数据量),需要将MySQL中数据全量/增量快速迁移至DBLE。TiDB DM支持多线程数据导出导入及增量同步,因此选择了TiDB DM为作数据迁移工具。本文以此案例为背景,介绍使用TiDB DM将数据从MySQL迁移至DBLE的方法及遇到的一些问题。
二、数据迁移示意图及服务器环境
(1)数据迁移示意图如下:
(2)服务器环境如下:
服务器IP | 角色 | 端口 | 版本 | 备注 |
---|---|---|---|---|
10.186.65.83 | dm master | 8261/8291 | V2.0.7 | tiup安装节点 |
10.186.65.118 | dm worker | 8262 | V2.0.7 | |
10.186.65.14 | SRC MySQL | 4490 | MySQL 8.0.18 | |
10.186.65.4 | DBLE | 8066 | DBLE 3.21.10.6 | |
10.186.65.61 | DBLE datanode | 4408 | MySQL 8.0.25 | |
10.186.65.65 | DBLE datanode | 4408 | MySQL 8.0.25 |
三、安装 TiDB DM
(1)准备环境及安装 tiup
#分别在dm master和worker节点主机创建tidb用户,密码为dmadmin,并配置sudo
shell> useradd tidb
shell> echo "dmadmin" | passwd tidb --stdin
shell> echo "tidb ALL=(ALL) NOPASSWD: ALL" > /etc/sudoers.d/tidb
#切换至tidb用户,生成密钥,并做ssh信任,dm集群内机器都需要做ssh信任
shell> su - tidb
shell> ssh-keygen -t rsa
shell> ssh-copy-id -i ~/.ssh/id_rsa.pub [email protected]
shell> ssh-copy-id -i ~/.ssh/id_rsa.pub [email protected]
#安装tiup,在tidb用户下
shell> curl --proto '=https' --tlsv1.2 -sSf https://tiup-mirrors.pingcap.com/install.sh | sh
shell> source .bash_profile
shell> which tiup
~/.tiup/bin/tiup
#安装dmctl
shell> tiup install dm dmctl:v2.0.7
component dm version v1.10.2 is already installed
download https://tiup-mirrors.pingcap.com/dmctl-v2.0.7-linux-amd64.tar.gz 26.92 MiB / 26.92 MiB 100.00% 10.01 MiB/s
(2)安装dm-cluster
#生成配置模版
shell> tiup dm template > dm_topology.yaml
#修改后配置文件如下:
shell> cat dm_topology.yaml |grep -v ^[#-] |grep -v ^$
global:
user: "tidb"
ssh_port: 22
deploy_dir: "/home/tidb/dm/deploy"
data_dir: "/home/tidb/dm/data"
master_servers:
- host: 10.186.65.83
worker_servers:
- host: 10.186.65.118
monitoring_servers:
- host: 10.186.65.83
grafana_servers:
- host: 10.186.65.83
alertmanager_servers:
- host: 10.186.65.83
#安装、并启动DM集群
shell> tiup dm deploy dm-test v2.0.7 ./dm_topology.yaml --user tidb -i /home/tidb/.ssh/id_rsa
shell> tiup dm start dm-test
#查看DM集群
[tidb@10_186_65_83 ~]$ tiup dm display dm-test
tiup is checking updates for component dm ...
Starting component `dm`: /home/tidb/.tiup/components/dm/v1.10.2/tiup-dm display dm-test
Cluster type: dm
Cluster name: dm-test
Cluster version: v2.0.7
Deploy user: tidb
SSH type: builtin
Grafana URL: http://10.186.65.83:3000
ID Role Host Ports OS/Arch Status Data Dir Deploy Dir
-- ---- ---- ----- ------- ------ -------- ----------
10.186.65.83:9093 alertmanager 10.186.65.83 9093/9094 linux/x86_64 Up /home/tidb/dm/data/alertmanager-9093 /home/tidb/dm/deploy/alertmanager-9093
10.186.65.83:8261 dm-master 10.186.65.83 8261/8291 linux/x86_64 Healthy|L /home/tidb/dm/data/dm-master-8261 /home/tidb/dm/deploy/dm-master-8261
10.186.65.118:8262 dm-worker 10.186.65.118 8262 linux/x86_64 Free /home/tidb/dm/data/dm-worker-8262 /home/tidb/dm/deploy/dm-worker-8262
10.186.65.83:3000 grafana 10.186.65.83 3000 linux/x86_64 Up - /home/tidb/dm/deploy/grafana-3000
10.186.65.83:9090 prometheus 10.186.65.83 9090 linux/x86_64 Up /home/tidb/dm/data/prometheus-9090 /home/tidb/dm/deploy/prometheus-9090
Total nodes: 5
四、DBLE端配置
(1)db.xml
select 1
select 1
#注:用户密码为加密后的字符串
(2)sharding.xml
4
0:0
注:需要提前配置好DM同步时保存数据的schema及两张表,表名分别是
dm同步任务名 + _loader_checkpoint 即表mysql_sync_to_dble_loader_checkpoint
dm同步任务名 + _syncer_checkpoint 即表mysql_sync_to_dble_syncer_checkpoint
(3)user.xml
#注:用户密码为加密后的字符串
(4)登陆 DBLE 管理端执行创建物理数据库命令
shell> /data/mysql/base/5.7.36/bin/mysql -uroot -padmin -h 10.186.65.4 -P 9066
dble> create database @@shardingNode = 'dn$1-4';
Query OK, 1 row affected (0.06 sec)
dble> create database @@shardingNode = 'dm_meta';
Query OK, 1 row affected (0.01 sec)
五、在源端 MySQL 压测创建测试数据
#执行prepare,初始化100W行数据
/usr/share/sysbench/oltp_insert.lua --mysql-db=sbtest --mysql-host=10.186.65.14 --mysql-port=4490 --mysql-user=sgy --mysql-password=admin --tables=2 --table-size=1000000 --threads=16 --time=60 --report-interval=1 --max-requests=0 --percentile=95 --mysql-ignore-errors=1062 prepare
sysbench 1.0.17 (using system LuaJIT 2.0.4)
Initializing worker threads...
Creating table 'sbtest1'...
Creating table 'sbtest2'...
Inserting 1000000 records into 'sbtest2'
Inserting 1000000 records into 'sbtest1'
Creating a secondary index on 'sbtest2'...
Creating a secondary index on 'sbtest1'...
#利用sysbench不断写入数据
shell> /usr/share/sysbench/oltp_read_write.lua --mysql-db=sbtest --mysql-host=10.186.65.14 --mysql-port=4490 --mysql-user=sgy --mysql-password=admin --tables=2 --table-size=2000000 --threads=16 --time=6000 --report-interval=1 --max-requests=0 --percentile=95 --mysql-ignore-errors=1062 run
sysbench 1.0.17 (using system LuaJIT 2.0.4)
Running the test with following options:
Number of threads: 16
Report intermediate results every 1 second(s)
Initializing random number generator from current time
Initializing worker threads...
Threads started!
[ 1s ] thds: 16 tps: 585.44 qps: 11916.49 (r/w/o: 8360.33/1282.63/2273.53) lat (ms,95%): 55.82 err/s: 0.00 reconn/s: 0.00
[ 2s ] thds: 16 tps: 744.18 qps: 14819.02 (r/w/o: 10379.16/1609.36/2830.50) lat (ms,95%): 41.85 err/s: 0.00 reconn/s: 0.00
[ 3s ] thds: 16 tps: 605.11 qps: 12144.20 (r/w/o: 8500.54/1321.24/2322.42) lat (ms,95%): 47.47 err/s: 0.00 reconn/s: 0.00
[ 4s ] thds: 16 tps: 773.91 qps: 15558.16 (r/w/o: 10881.71/1737.79/2938.65) lat (ms,95%): 41.10 err/s: 0.00 reconn/s: 0.00
六、启动数据同步任务
(1)配置数据源
#使用dmctl工具生成密码加密字符串,每次执行都会产生不同的加密字符串
shell> tiup dmctl encrypt 'admin'
dmXgktpuF18RP1mDN/B2UkA6pDN9PQ==
#mysql数据库配置文件如下
shell> cat mysql_source_14_4490.yaml
source-id: "mysql_source_14_4490"
from:
host: "10.186.65.14"
port: 4490
user: "sgy"
password: "dmXgktpuF18RP1mDN/B2UkA6pDN9PQ=="
#创建数据源
shell> tiup dmctl --master-addr 10.186.65.83:8261 operate-source create ./mysql_source_14_4490.yaml
Starting component `dmctl`: /home/tidb/.tiup/components/dmctl/v2.0.7/dmctl/dmctl --master-addr 10.186.65.83:8261 operate-source create ./mysql_source_14_4490.yaml
{
"result": true,
"msg": "",
"sources": [
{
"result": true,
"msg": "",
"source": "mysql_source_14_4490",
"worker": "dm-10.186.65.118-8262"
}
]
}
#查看数据源
shell> tiup dmctl --master-addr 10.186.65.83:8261 operate-source show
{
"result": true,
"msg": "",
"sources": [
{
"result": true,
"msg": "",
"source": "mysql_source_14_4490",
"worker": "dm-10.186.65.118-8262"
}
]
}
(2)配置同步任务
#同步配置文件如下
shell> cat mysql_sync_to_dble.yaml
name: "mysql_sync_to_dble" #同步任务名
task-mode: "all" #全量 + Binlog 实时同步
clean-dump-file: true
timezone: "Asia/Shanghai"
ignore-checking-items: ["auto_increment_ID"]
target-database:
host: "10.186.65.4" #DBLE主机IP地址
port: 8066 #DBLE流量端口
user: "sz" #连接DBLE的用户
password: "KRfSNtdxe63tehpnCYoCz0ABdUGivg==" #连接DBLE用户密码,用tiup dmctl encrypt 'password'生成
mysql-instances:
-
source-id: "mysql_source_14_4490"
block-allow-list: "global"
mydumper-config-name: "global"
loader-config-name: "global"
syncer-config-name: "global"
block-allow-list:
global:
do-dbs: ["sbtest"] #只迁移sbtest库
mydumpers:
global:
extra-args: "-B sbtest" #只dump sbtest库
loaders:
global:
pool-size: 8
syncers:
global:
worker-count: 8
#更详细配置说明参考官方文档:
https://docs.pingcap.com/zh/tidb-data-migration/v2.0/task-configuration-file-full#%E5%AE%8C%E6%95%B4%E9%85%8D%E7%BD%AE%E6%96%87%E4%BB%B6%E7%A4%BA%E4%BE%8B
#检查配置文件
shell> tiup dmctl:v2.0.7 --master-addr=10.186.65.83:8261 check-task ./mysql_sync_to_dble.yaml
Starting component `dmctl`: /home/tidb/.tiup/components/dmctl/v2.0.7/dmctl/dmctl --master-addr=10.186.65.83:8261 check-task ./mysql_sync_to_dble.yaml
{
"result": true,
"msg": "check pass!!!"
}
#启动同步任务
shell> tiup dmctl:v2.0.7 --master-addr=10.186.65.83:8261 start-task ./mysql_sync_to_dble.yaml
Starting component `dmctl`: /home/tidb/.tiup/components/dmctl/v2.0.7/dmctl/dmctl --master-addr=10.186.65.83:8261 start-task ./mysql_sync_to_dble.yaml
{
"result": true,
"msg": "",
"sources": [
{
"result": true,
"msg": "",
"source": "mysql_source_14_4490",
"worker": "dm-10.186.65.118-8262"
}
]
}
#在dm-worker节点查看导出的SQL文件,
shell> ls -lh /home/tidb/dm/deploy/dm-worker-8262/dumped_data.mysql_sync_to_dble
total 384M
-rw-r--r-- 1 tidb tidb 187 Aug 5 14:04 metadata
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest1.0000000000000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest1.0000000010000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest1.0000000020000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest1.0000000030000.sql
-rw-r--r-- 1 tidb tidb 369 Aug 5 14:04 sbtest.sbtest1-schema.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest2.0000000000000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest2.0000000010000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest2.0000000020000.sql
-rw-r--r-- 1 tidb tidb 48M Aug 5 14:04 sbtest.sbtest2.0000000030000.sql
-rw-r--r-- 1 tidb tidb 369 Aug 5 14:04 sbtest.sbtest2-schema.sql
-rw-r--r-- 1 tidb tidb 152 Aug 5 14:04 sbtest-schema-create.sql
#创建库SQL文件
shell> head /home/tidb/dm/deploy/dm-worker-8262/dumped_data.mysql_sync_to_dble/sbtest-schema-create.sql
/*!40101 SET NAMES binary*/;
CREATE DATABASE `sbtest` /*!40100 DEFAULT CHARACTER SET utf8mb4 COLLATE utf8mb4_bin */ /*!80016 DEFAULT ENCRYPTION='N' */;
#创建表SQL文件
shell> head /home/tidb/dm/deploy/dm-worker-8262/dumped_data.mysql_sync_to_dble/sbtest.sbtest1-schema.sql
/*!40101 SET NAMES binary*/;
CREATE TABLE `sbtest1` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`k` int(11) NOT NULL DEFAULT '0',
`c` char(120) COLLATE utf8mb4_bin NOT NULL DEFAULT '',
`pad` char(60) COLLATE utf8mb4_bin NOT NULL DEFAULT '',
PRIMARY KEY (`id`),
KEY `k_1` (`k`)
) ENGINE=InnoDB AUTO_INCREMENT=1000001 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin;
#dump生成批量插入数据SQL文件
shell> head /home/tidb/dm/deploy/dm-worker-8262/dumped_data.mysql_sync_to_dble/sbtest.sbtest1.0000000000000.sql
/*!40101 SET NAMES binary*/;
INSERT INTO `sbtest1` (`id`,`k`,`c`,`pad`) VALUES
(1,498670,'31451373586-15688153734-79729593694-96509299839-83724898275-86711833539-78981337422-35049690573-51724173961-87474696253','98996621624-36689827414-04092488557-09587706818-65008859162'),
(2,497778,'21472970079-70972780322-70018558993-71769650003-09270326047-32417012031-10768856803-14235120402-93989080412-18690312264','04776826683-45880822084-77922711547-29057964468-76514263618'),
(3,498956,'49376827441-24903985029-56844662308-79012577859-40518387141-60588419212-24399130405-42612257832-29494881732-71506024440','26843035807-96849339132-53943793991-69741192222-48634174017'),
(4,518727,'85762858421-36258200885-10758669419-44272723583-12529521893-95630803635-53907705724-07005352902-43001596772-53048338959','37979424284-37912826784-31868864947-42903702727-96097885121'),
(5,502480,'24805466175-85245528617-94635882649-46305216925-28637832581-03224489581-68883711727-95491561683-91969681472-12022277774','19288959552-55556468076-14192290426-55457672510-18043372364'),
(6,500774,'52892836230-54177743992-01821871718-48412537487-30066596248-87215430797-00375777469-64498831720-58542556455-90784765418','59487960480-08453890592-99628797439-16757639138-29377916560'),
(7,501466,'85820931248-14475640036-11980694501-86588543167-31029306229-09626867980-90685354565-02350460358-25863585366-53793794448','26081374730-86321700986-51212137094-30635959762-03880194434'),
(8,497838,'81578049255-33453976301-67096870761-27658738403-30546242249-53677469854-26594573136-34292002037-52736825353-99165193170','64289062455-51067794311-09919261228-11533354367-07401173317'),
(3)查看同步任务状态
shell> tiup dmctl:v2.0.7 --master-addr=10.186.65.83:8261 query-status ./mysql_sync_to_dble.yaml
Starting component `dmctl`: /home/tidb/.tiup/components/dmctl/v2.0.7/dmctl/dmctl --master-addr=10.186.65.83:8261 query-status ./mysql_sync_to_dble.yaml
{
"result": true,
"msg": "",
"sources": [
{
"result": true,
"msg": "",
"sourceStatus": {
"source": "mysql_source_14_4490",
"worker": "dm-10.186.65.118-8262",
"result": null,
"relayStatus": null
},
"subTaskStatus": [
{
"name": "mysql_sync_to_dble",
"stage": "Running",
"unit": "Sync",
"result": null,
"unresolvedDDLLockID": "",
"sync": {
"totalEvents": "425355",
"totalTps": "1203",
"recentTps": "2386",
"masterBinlog": "(mysql-bin.000027, 76114275)",
"masterBinlogGtid": "c7827165-bf89-11ec-92e6-02000aba410e:1-5091258",
"syncerBinlog": "(mysql-bin.000026, 229462770)",
"syncerBinlogGtid": "",
"blockingDDLs": [
],
"unresolvedGroups": [
],
"synced": false,
"binlogType": "remote",
"secondsBehindMaster": "38"
}
}
]
}
]
}
七、验证数据
#暂停先前执行的sysbench程序,确保dm同步与源端无延迟时,对表作count(*)
#在源端MySQL,对表进行count(*)
mysql> select count(*) from sbtest.sbtest1;
+----------+
| count(*) |
+----------+
| 1116471 |
+----------+
1 row in set (0.68 sec)
mysql> select count(*) from sbtest.sbtest2;
+----------+
| count(*) |
+----------+
| 1117020 |
+----------+
1 row in set (0.54 sec)
#通过dble对表进行count(*)
dble> select count(*) from sbtest.sbtest1;
+----------+
| count(*) |
+----------+
| 1116471 |
+----------+
1 row in set (1.04 sec)
dble> select count(*) from sbtest.sbtest2;
+----------+
| count(*) |
+----------+
| 1117020 |
+----------+
1 row in set (1.58 sec)
#在源端数据进行更新操作
mysql> update sbtest.sbtest1 set c=uuid() where id=20;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1 Changed: 1 Warnings: 0
mysql> select * from sbtest.sbtest1 where id=20;
+----+--------+--------------------------------------+-------------------------------------------------------------+
| id | k | c | pad |
+----+--------+--------------------------------------+-------------------------------------------------------------+
| 20 | 501448 | 24649f0f-14d1-11ed-b4f2-02000aba410e | 91052688950-96415657187-00012408429-12357288330-41295735957 |
+----+--------+--------------------------------------+-------------------------------------------------------------+
1 row in set (0.00 sec)
#通过dble查询数据
dble> select * from sbtest.sbtest1 where id=20;
+----+--------+--------------------------------------+-------------------------------------------------------------+
| id | k | c | pad |
+----+--------+--------------------------------------+-------------------------------------------------------------+
| 20 | 501448 | 24649f0f-14d1-11ed-b4f2-02000aba410e | 91052688950-96415657187-00012408429-12357288330-41295735957 |
+----+--------+--------------------------------------+-------------------------------------------------------------+
1 row in set (0.01 sec)
八、注意事项
(1)TiDB DM版本要求:TiDB DM v2.0.7
(2)在进行数据量导入时需要关闭DBLE慢查询日志
#由于数据同步在全量导入阶段会产生大量慢查询,需要在执行同步任务之前先关闭DBLE的慢查询日志
shell> mysql --prompt='dble>' -uroot -p -h 10.186.65.4 -P 9066
#查看慢查询日志状态,1表示开启,0表示关闭
dble> show @@slow_query_log;
+------------------+
| @@slow_query_log |
+------------------+
| 1 |
+------------------+
1 row in set (0.00 sec)
#全量数据导入前,关闭dble慢查询日志
dble> disable @@slow_query_log;
Query OK, 1 row affected (0.01 sec)
disable slow_query_log success
#待全量数据导入完成后,开启dble慢查询日志
dble> enable @@slow_query_log;
Query OK, 1 row affected (0.01 sec)
enable slow_query_log success
(3)检查源端 MySQL 实例上是否有运行类似 pt-kill 的程序
#如果在源端MySQL实例上运行有pt-kill或类似的程序,DM同步任务在执行全量数据dump阶段,
导出线程可能会被程序kill掉,导致全备失败。dm-worker.log日志如下:
[ERROR] [subtask.go:311] ["unit process error"] [subtask=oms-sync-receiver] [unit=Dump] ["error information"="{\"ErrCode\":32001,\"ErrClass\":\"dump-unit\",\"ErrScope\":\"internal\",\"ErrLevel\":\"high\",\"Message\":\"mydumper/dumpling runs with error, with output (may empty): \",\"RawCause\":\"invalid connection\"}"]
解决方法:先将pt-kill实用工具停止,然后再重新运行TiDM同步任务,待全量数据dump完成后在开启pt-kill。
(4)迁移时如需对目标表表结构进行修改,如源表有建立分区而目标表通过分片后不需要建立分区、修改表字符集(建议使用utf8mb4)等,可先将表结构从源库导出,修改后导入 DBLE 或导入 DBLE 后在修改表结构,最后再开启 DM 同步任务。
(5)源、目标表字段顺序必须要一致,否则可能会导致数据不一致,两字段同为 varchar 并且长度不相等时,同步任务并不会报错,但写入目标表的值可能会被截断,如下图:
DM 任务启动后连接目标端时,会将 session 级别 sql_mode 设置成下面值:
sql_mode='IGNORE_SPACE,NO_AUTO_VALUE_ON_ZERO,NO_ENGINE_SUBSTITUTION,ALLOW_INVALID_DATES,ONLY_FULL_GROUP_BY';