-
Notifications
You must be signed in to change notification settings - Fork 0
HA配置
注意： 两台机器上的instance目录的名字需要保证完全一致，HA模式是依赖于instance name进行管理，同时必须都选择default-instance.xml配置
- mysql master ：192.168.74.129 端口 3306
- 两个canal ：192.168.74.129、 192.168.74.131
- zookeeper ：192.168.74.129:2181
分别配置canal文件夹conf下的
canal.properties和instance文件夹下的instance.properties保持两个canal的instance name相同
a. 修改canal.properties，加上zookeeper配置
canal.zkServers=192.168.74.129:2181
canal.instance.global.spring.xml = classpath:spring/default-instance.xml
b. 创建example目录(如果不存在example文件夹)，并修改instance.properties
canal.instance.mysql.slaveId = 1234 ##另外一台机器改成1235，保证slaveId不重复即可
canal.instance.master.address = 192.168.74.129:3306 ## mysql
canal配置参考如下：
- 启动mysql
- 启动zookeeper集群
- 启动两台canal
启动 192.168.74.129
[root@localhost canal]# sh bin/startup.sh
启动 192.168.74.131
[root@localhost canal]# sh bin/startup.sh
查看启动日志发现只会有一台canal启动（192.168.74.129启动日志）
[root@yyp canal]# tail -1000f logs/example/example.log
2017-07-15 23:06:14.842 [main] INFO c.a.o.c.i.spring.support.PropertyPlaceholderConfigurer - Loading properties file from class path resource [canal.pro
perties]2017-07-15 23:06:14.849 [main] INFO c.a.o.c.i.spring.support.PropertyPlaceholderConfigurer - Loading properties file from class path resource [example/i
nstance.properties]2017-07-15 23:06:14.984 [main] WARN org.springframework.beans.TypeConverterDelegate - PropertyEditor [com.sun.beans.editors.EnumEditor] found through de
precated global PropertyEditorManager fallback - consider using a more isolated form of registration, e.g. on the BeanWrapper/BeanFactory!2017-07-15 23:06:15.237 [main] INFO c.a.otter.canal.instance.spring.CanalInstanceWithSpring - start CannalInstance for 1-example
2017-07-15 23:06:15.277 [main] INFO c.a.otter.canal.instance.core.AbstractCanalInstance - start successful....
2017-07-15 23:06:15.426 [destination = example , address = /192.168.74.129:3306 , EventParser] WARN c.a.otter.canal.parse.inbound.mysql.MysqlEventParser
- prepare to find start position just show master status
- zookeeper查看当前运行的节点信息
登陆zookeeper(192.168.74.129:2181)，切到bin目录，执行
zkCli.cmd 或zkCli.sh命令
再执行命令查看节点状态
[zk: localhost:2181(CONNECTED) 2] get /otter/canal/destinations/example/running
{"active":true,"address":"192.168.74.129:11111","cid":1}
cZxid = 0x20000003a
ctime = Sat Jul 15 23:21:36 EDT 2017
mZxid = 0x20000003a
mtime = Sat Jul 15 23:21:36 EDT 2017
pZxid = 0x20000003a
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x15d495b78580003
dataLength = 56
numChildren = 0
[zk: localhost:2181(CONNECTED) 3]
a. 可以直接指定zookeeper地址和instance name，canal client会自动从zookeeper中的running节点，获取当前服务的工作节点，然后与其建立链接：(此处使用集群client测试：com.alibaba.otter.canal.example.ClusterCanalClientTest)
String destination = "example";
// 基于固定canal server的地址，建立链接，其中一台server发生crash，可以支持failover
// CanalConnector connector = CanalConnectors.newClusterConnector(
// Arrays.asList(new InetSocketAddress(
// AddressUtils.getHostIp(),
// 11111)),
// "stability_test", "", "");
// 基于zookeeper动态获取canal server的地址，建立链接，其中一台server发生crash，可以支持failover
CanalConnector connector = CanalConnectors.newClusterConnector("192.168.74.129:2181", destination, "", "");b. 链接成功后，canal server会记录当前正在工作的canal client信息，比如客户端ip，链接的端口信息等
[zk: localhost:2181(CONNECTED) 2] get /otter/canal/destinations/example/1001/running
{"active":true,"address":"192.168.74.1:32846","clientId":1001}
c. 登陆mysql mater，插入一条数据
mysql> insert into person values(5,'use zookeeper HA');
Query OK, 1 row affected (0.60 sec)
mysql> d. 数据消费成功后，canal server会在zookeeper中记录下当前最后一次消费成功的binlog位点. (下次你重启client时，会从这最后一个位点继续进行消费)
[zk: localhost:2181(CONNECTED) 3] get /otter/canal/destinations/example/1001/cursor
{"@type":"com.alibaba.otter.canal.protocol.position.LogPosition","identity":{"slaveId":-1,"sourceAddress":{"address":"192.168.74.129","port":3306}},"post
ion":{"included":false,"journalName":"mysql-binlog.000002","position":3985,"serverId":1,"timestamp":1500176724000}}
e. 查看mysql slave同步成功，并且idel同时输出信息：
****************************************************
* Batch Id: [1] ,count : [3] , memsize : [164] , Time : 2017-07-16 11:39:24
* Start : [mysql-binlog.000002:3797:1500176724000(2017-07-16 11:45:24)]
* End : [mysql-binlog.000002:3985:1500176724000(2017-07-16 11:45:24)]
****************************************************
================> binlog[mysql-binlog.000002:3797] , executeTime : 1500176724000 , delay : -359418ms
BEGIN ----> Thread id: 51
----------------> binlog[mysql-binlog.000002:3927] , name[test_db,person] , eventType : INSERT , executeTime : 1500176724000 , delay : -359397ms
pId : 5 type=int(11) update=true
pName : use zookeeper HA type=varchar(100) update=true
----------------
END ----> transaction id: 389
================> binlog[mysql-binlog.000002:3985] , executeTime : 1500176724000 , delay : -359293ms
停止 192.168.74.129 canal
[root@yyp canal]# sh bin/stop.sh
192.168.74.129 canal日志停止成功
2017-07-15 23:10:11.084 [Thread-4] INFO c.a.otter.canal.instance.core.AbstractCanalInstance - stop CannalInstance for null-example
2017-07-15 23:10:11.107 [Thread-4] INFO c.a.otter.canal.instance.core.AbstractCanalInstance - stop successful....
此时登陆 192.168.74.131 canal，查看日志
2017-07-15 23:10:23.730 [pool-5-thread-1] INFO c.a.o.c.i.spring.support.PropertyPlaceholderConfigurer - Lo
ading properties file from class path resource [canal.properties]2017-07-15 23:10:23.743 [pool-5-thread-1] INFO c.a.o.c.i.spring.support.PropertyPlaceholderConfigurer - Lo
ading properties file from class path resource [example/instance.properties]2017-07-15 23:10:23.747 [pool-5-thread-1] WARN org.springframework.beans.TypeConverterDelegate - PropertyE
ditor [com.sun.beans.editors.EnumEditor] found through deprecated global PropertyEditorManager fallback - consider using a more isolated form of registration, e.g. on the BeanWrapper/BeanFactory!2017-07-15 23:10:23.750 [pool-5-thread-1] INFO c.a.otter.canal.instance.spring.CanalInstanceWithSpring - s
tart CannalInstance for 1-example 2017-07-15 23:10:23.782 [pool-5-thread-1] INFO c.a.otter.canal.instance.core.AbstractCanalInstance - start
successful....2017-07-15 23:10:23.862 [destination = example , address = /192.168.74.129:3306 , EventParser] WARN c.a.ot
ter.canal.parse.inbound.mysql.MysqlEventParser - prepare to find start position just show master status
- HA验证结果：
- 这时192.168.74.131会立马启动example instance，提供新的数据服务
- 客户端也会随着canal server的切换，通过获取zookeeper中的最新地址，与新的canal server建立链接，继续消费数据，整个过程自动完成*
-
正常场景
a. 正常关闭canal server(会释放instance的所有资源，包括删除running节点)
b. 平滑切换(gracefully)
操作：更新对应instance的running节点内容，将"active"设置为false，对应的running节点收到消息后，会主动释放running节点，让出控制权但自己jvm不退出，gracefully.
-
异常场景
a. canal server对应的jvm异常crash，running节点的释放会在对应的zookeeper session失效后，释放running节点(EPHEMERAL节点) ps. session过期时间默认为zookeeper配置文件中定义的tickTime的20倍，如果不改动zookeeper配置，那默认就是40秒
b. canal server所在的网络出现闪断，导致zookeeper认为session失效，释放了running节点，此时canal server对应的jvm并未退出，(一种假死状态，非常特殊的情况)
ps. 为了保护假死状态的canal server，避免因瞬间runing失效导致instance重新分布，所以做了一个策略：canal server在收到running节点释放后，延迟一段时间抢占running，原本running节点的拥有者可以不需要等待延迟，优先取得running节点，可以保证假死状态下尽可能不无谓的释放资源。 目前延迟时间的默认值为5秒，即running节点针对假死状态的保护期为5秒.