canal 是阿里巴巴开源的一个项目,主要用途是基于 MySQL 数据库 binlog 日志解析,提供增量数据订阅和消费。
基于日志增量订阅和消费的业务包括:
- 数据库镜像
- 数据库实时备份
- 索引构建和实时维护(拆分异构索引、倒排索引等)
- 业务 cache 刷新
- 带业务逻辑的增量数据处理
我这边主要在两个场景下使用:
一个是将变更数据实时同步到 Elasticsearch 和 Redis。
这里先说一下我目前的做法,一方面是全量数据定时同步,由于数据量比较大,同步时间比较长,所以数据也就不够实时。第二个方面是针对单条数据的变更,部分更新 Elasticsearch 和 Redis 的逻辑都是直接写在了业务代码中,耦合比较严重。
拆出来之后就可以实现实时增量更新,而且还可以解耦,收益还是很大的。
第二个是保存重点关注数据的历史变更。
这个目前用在了「资产管理」模块,通过记录 IP 资产的创建,变更以及删除,实现 IP 生命周期管理,方便历史信息回溯。
MySQL 配置
修改 MySQL 配置文件 my.cnf,开启 binlog 写入功能,并配置模式为 ROW。
log-bin=mysql-bin # 开启 binlog
binlog-format=ROW # 选择 ROW 模式
server_id=1 # 配置 MySQL replaction 需要定义,不要和 canal 的 slaveId 重复
重启数据库,查看配置是否生效。
mysql> show variables like 'binlog_format';
+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| binlog_format | ROW |
+---------------+-------+
1 row in set (0.19 sec)
mysql>
mysql> show variables like 'log_bin';
+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| log_bin | ON |
+---------------+-------+
1 row in set (0.00 sec)
mysql>
mysql> show master status;
+------------------+----------+--------------+------------------+-------------------+
| File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set |
+------------------+----------+--------------+------------------+-------------------+
| mysql-bin.000003 | 4230 | | | |
+------------------+----------+--------------+------------------+-------------------+
1 row in set (0.00 sec)
然后创建用户,并授权。
mysql> CREATE USER canal IDENTIFIED BY 'canal';
mysql> GRANT SELECT, REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO 'canal'@'%%';
mysql> FLUSH PRIVILEGES;
mysql> show grants for 'canal'@'%%';
+----------------------------------------------------------------------------+
| Grants for canal@%% |
+----------------------------------------------------------------------------+
| GRANT SELECT, REPLICATION SLAVE, REPLICATION CLIENT ON *.* TO `canal`@`%%` |
+----------------------------------------------------------------------------+
1 row in set (0.00 sec)
canal 服务端
拉取镜像:
# docker pull canal/canal-server:v1.1.4
然后用官方提供的 shell 脚本直接启动:
# sh run.sh -e canal.auto.scan=false -e canal.destinations=test -e canal.instance.master.address=127.0.0.1:3306 -e canal.instance.dbUsername=canal -e canal.instance.dbPassword=canal -e canal.instance.connectionCharset=UTF-8 -e canal.instance.tsdb.enable=true -e canal.instance.gtidon=false
但每次都这样启动还是有点麻烦,可以写一个 docker-compose 文件,如下:
version: '3'
services:
canal-server:
image: canal/canal-server:v1.1.4
container_name: canal-server
restart: unless-stopped
network_mode: host
ports:
- 11111:11111
environment:
- canal.auto.scan=false
- canal.instance.master.address=127.0.0.1:3306
- canal.instance.dbUsername=canal
- canal.instance.dbPassword=canal
- canal.instance.filter.regex=.*\\..*
- canal.destinations=test
- canal.instance.connectionCharset=UTF-8
- canal.instance.tsdb.enable=true
volumes:
- /root/canal/test/log/:/home/admin/canal-server/logs/
启动服务:
# docker-compose up
Recreating canal-server ... done
Attaching to canal-server
canal-server | DOCKER_DEPLOY_TYPE=VM
canal-server | ==> INIT /alidata/init/02init-sshd.sh
canal-server | ==> EXIT CODE: 0
canal-server | ==> INIT /alidata/init/fix-hosts.py
canal-server | ==> EXIT CODE: 0
canal-server | ==> INIT DEFAULT
canal-server | Generating SSH1 RSA host key: [ OK ]
canal-server | Starting sshd: [ OK ]
canal-server | Starting crond: [ OK ]
canal-server | ==> INIT DONE
canal-server | ==> RUN /home/admin/app.sh
canal-server | ==> START ...
canal-server | start canal ...
canal-server | start canal successful
canal-server | ==> START SUCCESSFUL ...
canal Python 客户端
直接 Copy 官方提供的客户端代码:
import time
from canal.client import Client
from canal.protocol import EntryProtocol_pb2
from canal.protocol import CanalProtocol_pb2
client = Client()
client.connect(host='127.0.0.1', port=11111)
client.check_valid(username=b'', password=b'')
client.subscribe(client_id=b'1001', destination=b'test', filter=b'.*\\..*')
while True:
message = client.get(100)
entries = message['entries']
for entry in entries:
entry_type = entry.entryType
if entry_type in [EntryProtocol_pb2.EntryType.TRANSACTIONBEGIN, EntryProtocol_pb2.EntryType.TRANSACTIONEND]:
continue
row_change = EntryProtocol_pb2.RowChange()
row_change.MergeFromString(entry.storeValue)
event_type = row_change.eventType
header = entry.header
database = header.schemaName
table = header.tableName
event_type = header.eventType
for row in row_change.rowDatas:
format_data = dict()
if event_type == EntryProtocol_pb2.EventType.DELETE:
for column in row.beforeColumns:
format_data = {
column.name: column.value
}
elif event_type == EntryProtocol_pb2.EventType.INSERT:
for column in row.afterColumns:
format_data = {
column.name: column.value
}
else:
format_data['before'] = format_data['after'] = dict()
for column in row.beforeColumns:
format_data['before'][column.name] = column.value
for column in row.afterColumns:
format_data['after'][column.name] = column.value
data = dict(
db=database,
table=table,
event_type=event_type,
data=format_data,
)
print(data)
time.sleep(1)
client.disconnect()
功能验证
首先在 MySQL 里边创建一张测试表,然后再增删改几条测试数据:
mysql> create database test;
mysql> use test;
mysql> CREATE TABLE `role` ( `id` int unsigned NOT NULL AUTO_INCREMENT, `role_name` varchar(255)
DEFAULT NULL, PRIMARY KEY (`id`) ) ENGINE=InnoDB AUTO_INCREMENT=2 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_0900_ai_ci;
mysql> insert into role (id, role_name) values (10, 'admin');
Query OK, 1 row affected (0.01 sec)
mysql> update role set role_name='hh' where id = 10;
Query OK, 1 row affected (0.01 sec)
Rows matched: 1 Changed: 1 Warnings: 0
mysql> delete from role where id = 10;
Query OK, 1 row affected (0.01 sec)
客户端打印输出:
$ python canal_client.py
connected to 127.0.0.1:11111
Auth succed
Subscribe succed
header {
version: 1
logfileName: "mysql-bin.000003"
logfileOffset: 5497
serverId: 1
serverenCode: "UTF-8"
executeTime: 1607843285000
sourceType: MYSQL
eventLength: 75
}
entryType: TRANSACTIONBEGIN
storeValue: " \217\001"
header {
version: 1
logfileName: "mysql-bin.000003"
logfileOffset: 5630
serverId: 1
serverenCode: "UTF-8"
executeTime: 1607843285000
sourceType: MYSQL
schemaName: "test"
tableName: "role"
eventLength: 47
eventType: INSERT
props {
key: "rowsCount"
value: "1"
}
}
entryType: ROWDATA
storeValue: "\010\322\001\020\001P\000bN\022 \010\000\020\004\032\002id \001(\0010\000B\00210R\014int unsigned\022*\010\001\020\014\032\trole_name \000(\0010\000B\005adminR\014varchar(255)"
{'db': 'test', 'table': 'role', 'event_type': 1, 'data': {'role_name': 'admin'}}
header {
version: 1
logfileName: "mysql-bin.000003"
logfileOffset: 5677
serverId: 1
serverenCode: "UTF-8"
executeTime: 1607843285000
sourceType: MYSQL
eventLength: 31
}
entryType: TRANSACTIONEND
storeValue: "\022\003440"
变更一条数据,输出内容分三部分,分别是:TRANSACTIONBEGIN,ROWDATA 和 TRANSACTIONEND。然后我们比较关注的内容都在 ROWDATA 中,解析出来之后就是我们需要的,包括数据库名,表名和变更内容。
其中 event_type 字段 1 表示新增,2 表示更新,3 表示删除。
update
对应输出:
{'db': 'test', 'table': 'role', 'event_type': 2, 'data': {'before': {'id': '10', 'role_name': 'hh'}, 'after': {'id': '10', 'role_name': 'hh'}}}
delete
对应输出:
{'db': 'test', 'table': 'role', 'event_type': 3, 'data': {'role_name': 'hh'}}
canal 服务端启动之后,在 /home/admin/canal-server/logs/test
目录下会生成两个日志文件,分别是:meta.log 和 test.log,可以查看服务是不是正常,有没有报错信息。其中 test 是启动 Docker 时 canal.destinations
设置的名称。
# cat meta.log
2020-12-13 14:55:18.051 - clientId:1001 cursor:[mysql-bin.000003,4805,1607842360000,1,] address[/127.0.0.1:3306]
2020-12-13 14:55:33.051 - clientId:1001 cursor:[mysql-bin.000003,5096,1607842531000,1,] address[127.0.0.1:3306]
2020-12-13 14:57:07.051 - clientId:1001 cursor:[mysql-bin.000003,5387,1607842625000,1,] address[127.0.0.1:3306]
# cat test.log
2020-12-13 14:55:09.067 [main] INFO c.a.otter.canal.instance.core.AbstractCanalInstance - start successful....
2020-12-13 14:55:09.144 [destination = test , address = /127.0.0.1:3306 , EventParser] WARN c.a.o.c.p.inbound.mysql.rds.RdsBinlogEventParserProxy - ---> begin to find start position, it will be long time for reset or first position
2020-12-13 14:55:09.144 [destination = test , address = /127.0.0.1:3306 , EventParser] WARN c.a.o.c.p.inbound.mysql.rds.RdsBinlogEventParserProxy - prepare to find start position just show master status
2020-12-13 14:55:09.693 [destination = test , address = /127.0.0.1:3306 , EventParser] WARN c.a.o.c.p.inbound.mysql.rds.RdsBinlogEventParserProxy - ---> find start position successfully, EntryPosition[included=false,journalName=mysql-bin.000003,position=4699,serverId=1,gtid=,timestamp=1607842360000] cost : 538ms , the next step is binlog dump
踩坑记录
在我自己搭建的测试环境一切正常,但放到项目 beta 环境上还是遇到了一个问题:
[fetch failed by table meta:
schemeName
.tableName
]
查了一下说是由于表删除,或者表结构变更引起的解析错误,增加一条配置就可以解决:
canal.instance.filter.table.error=true
加上之后,报错信息的确都没有了,但消费出来的数据没有 ROWDATA,这个地方确实困扰了我很长时间。
说实话,有的时候调试程序,并不怕碰到报错,怕的是没有报错,然后程序还不正常。
后来,我把忽略表错误的配置删除,又仔细看了一遍日志,发现还有一个报错:
Caused by: java.io.IOException: ErrorPacket [errorNumber=1142, fieldCount=-1, message=SHOW command denied to user
这明显就是权限不够嘛,问了一下我们的 DBA,果然如此,我们的 binlog 账号默认是没有 select 权限的,加上之后,问题就成功解决了。
静下心来仔细看日志是多么重要。
以上,下篇会说说对接 MQ 的事。
参考文档:
https://github.com/alibaba/canal
https://github.com/haozi3156666/canal-python