在至少有一个Leader存在的前提下，进行Zookeeper的在线增量、在线减量、在线迁移
在全过程中ZooKeeper不停止服务
注意事项
首先，当我们要从3台扩充到5台时，应保证集群不停止服务。
3台不停止服务的最低限度是2台（X/2+1），而5台的最低限度是3台。
-
我们应该保证，集群中最低有3台ZooKeeper是启动的。
-
此外，重启时应保证先重启
myid最小的机器，由小向大进行重启 -
Leader无论其myid大小，都放到最后重启
因为ZooKeeper的机制中，myid大的会向小的发起连接，而小的不会向大的发起连接。因此如果最后重启myid最小的机器，则其可能无法加入集群
环境情况
五台机器
| IP | Hostname |
|---|---|
| 10.1.24.110 | idc02-kafka-ds-00 |
| 10.1.24.111 | idc02-kafka-ds-01 |
| 10.1.24.112 | idc02-kafka-ds-02 |
| 10.1.24.113 | idc02-kafka-ds-03 |
| 10.1.24.114 | idc02-kafka-ds-04 |
JDK
jdk1.7.0_67
ZooKeeper
zookeeper-3.4.6
myid
根据IP自增为1-5
配置文件
|
1
2
3
4
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
|
实验过程
配置一个3节点的ZooKeeper
idc02-kafka-ds-00：
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-00bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
idc02-kafka-ds-01：
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-01 bin]$ ./zkServer.sh status
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: leader
|
idc02-kafka-ds-02：
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
将其扩容为5节点的ZooKeeper
先查看原先的ZooKeeper集群情况
echo mntr|nc localhost 2181
这条4字命令可以查看集群的情况，其中follower的相关数据需要在Leader机器上才能查看
在idc02-kafka-ds-01上查看
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
[hadoop@idc02-kafka-ds-01 bin]$ echo mntr|nc localhost 2181
zk_version 3.4.6-1569965, built on 02/20/2014 09:09 GMT
zk_avg_latency 0
zk_max_latency 0
zk_min_latency 0
zk_packets_received 3
zk_packets_sent 2
zk_num_alive_connections 1
zk_outstanding_requests 0
zk_server_state leader
zk_znode_count 4
zk_watch_count 0
zk_ephemerals_count 0
zk_approximate_data_size 27
zk_open_file_descriptor_count 27
zk_max_file_descriptor_count 65535
zk_followers 2
zk_synced_followers 2
zk_pending_syncs 0
|
启动另外两台机器的Zookeeper
另外两台机器的配置文件
|
1
2
3
4
5
6
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
server.4=10.1.24.113:2888:3888
server.5=10.1.24.114:2888:3888
|
启动
idc02-kafka-ds-03：
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-03 bin]# ./zkServer.sh status
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: follower
|
idc02-kafka-ds-04：
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-04bin]# ./zkServer.sh status
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
再查看集群情况
仍然在idc02-kafka-ds-01上查看
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
[hadoop@idc02-kafka-ds-01 bin]$ echo mntr|nc localhost 2181
zk_version 3.4.6-1569965, built on 02/20/2014 09:09 GMT
zk_avg_latency 0
zk_max_latency 0
zk_min_latency 0
zk_packets_received 4
zk_packets_sent 3
zk_num_alive_connections 1
zk_outstanding_requests 0
zk_server_state leader
zk_znode_count 4
zk_watch_count 0
zk_ephemerals_count 0
zk_approximate_data_size 27
zk_open_file_descriptor_count 31
zk_max_file_descriptor_count 65535
zk_followers 4
zk_synced_followers 4
zk_pending_syncs 0
|
可以看到zk_followers为4，连接到的follower从2变为4了
而且zk_synced_followers为4，说明新加入的2个也都同步好了
接下来我们滚动重启myid为1-3的前三台机器
先处理idc02-kafka-ds-00
关闭
如不放心请在关闭其间于
Leader机器或后加入的两台机器上监控日志
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-00bin]$./zkServer.shstop
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stoppingzookeeper...STOPPED
|
修改其配置文件
由原来的
|
1
2
3
4
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
|
到新的
|
1
2
3
4
5
6
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
server.4=10.1.24.113:2888:3888
server.5=10.1.24.114:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-00 bin]$ ./zkServer.sh start
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[hadoop@idc02-kafka-ds-00 bin]$ ./zkServer.sh status
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: follower
|
然后跳过作为Leader的idc02-kafka-ds-01，先处理idc02-kafka-ds-02
关闭
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstop
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stoppingzookeeper...STOPPED
|
修改配置文件
|
1
2
3
4
5
6
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
server.4=10.1.24.113:2888:3888
server.5=10.1.24.114:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstart
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Startingzookeeper...STARTED
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
最后处理原Leader的idc02-kafka-ds-01
关闭
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-01 bin]$ ./zkServer.sh stop
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stopping zookeeper ... STOPPED
|
查看新Leader
ZooKeeper会尽可能的选择myid最大的机器为Leader，因此原本的idc02-kafka-ds-04其myid为5变为了Leader
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-04bin]# ./zkServer.sh status
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:leader
|
修改配置文件
|
1
2
3
4
5
6
|
server.1=10.1.24.110:2888:3888
server.2=10.1.24.111:2888:3888
server.3=10.1.24.112:2888:3888
server.4=10.1.24.113:2888:3888
server.5=10.1.24.114:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-01bin]$./zkServer.shstart
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Startingzookeeper...STARTED
[hadoop@idc02-kafka-ds-01bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
在新的Leader上查看集群情况
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
[hadoop@idc02-kafka-ds-04 bin]# echo mntr|nc localhost 2181
zk_version 3.4.6-1569965, built on 02/20/2014 09:09 GMT
zk_avg_latency 1
zk_max_latency 4
zk_min_latency 0
zk_packets_received 12
zk_packets_sent 11
zk_num_alive_connections 1
zk_outstanding_requests 0
zk_server_state leader
zk_znode_count 4
zk_watch_count 0
zk_ephemerals_count 0
zk_approximate_data_size 27
zk_open_file_descriptor_count 33
zk_max_file_descriptor_count 65535
zk_followers 4
zk_synced_followers 4
zk_pending_syncs 0
|
一切正常
到这里，我们已经将原本的3台扩展到了5台，成功了一半。
然后只要将现在的5台再缩小到3台且不包括原本
myid为1-2的机器，就完成了迁移
将5台缩小回3台
修改idc02-kafka-ds-02
根据前面的注意事项，我们此时5台集群中启动的数量不得少于3台，因此我们需要先修改
3-5号机器的配置文件为3台，再关闭1-2号机器
关闭
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstop
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stoppingzookeeper...STOPPED
|
修改配置文件为
|
1
2
3
4
|
server.3=10.1.24.110:2888:3888
server.4=10.1.24.111:2888:3888
server.5=10.1.24.112:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstart
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Startingzookeeper...STARTED
[hadoop@idc02-kafka-ds-02bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
然后修改idc02-kafka-ds-03
关闭
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-03 bin]# ./zkServer.sh stop
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stopping zookeeper ... STOPPED
|
修改配置文件为
|
1
2
3
4
|
server.3=10.1.24.110:2888:3888
server.4=10.1.24.111:2888:3888
server.5=10.1.24.112:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-03 bin]$ ./zkServer.sh start
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[hadoop@idc02-kafka-ds-03 bin]$ ./zkServer.sh status
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode: follower
|
最后修改idc02-kafka-ds-04
关闭
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-04bin]$./zkServer.shstop
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stoppingzookeeper...STOPPED
|
关闭后
Leader移动到了myid第二大的idc02-kafka-ds-02上
修改配置文件为
|
1
2
3
4
|
server.3=10.1.24.110:2888:3888
server.4=10.1.24.111:2888:3888
server.5=10.1.24.112:2888:3888
|
启动
|
1
2
3
4
5
6
7
8
9
|
[hadoop@idc02-kafka-ds-04bin]$./zkServer.shstart
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Startingzookeeper...STARTED
[hadoop@idc02-kafka-ds-04bin]$./zkServer.shstatus
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Mode:follower
|
在Leader中查看
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
[hadoop@idc02-kafka-ds-03 bin]$ echo mntr|nc localhost 2181
zk_version 3.4.6-1569965, built on 02/20/2014 09:09 GMT
zk_avg_latency 0
zk_max_latency 0
zk_min_latency 0
zk_packets_received 4
zk_packets_sent 3
zk_num_alive_connections 1
zk_outstanding_requests 0
zk_server_state leader
zk_znode_count 4
zk_watch_count 0
zk_ephemerals_count 0
zk_approximate_data_size 27
zk_open_file_descriptor_count 27
zk_max_file_descriptor_count 65535
zk_followers 2
zk_synced_followers 2
zk_pending_syncs 0
|
此时的zk_followers为2，说明Leader已经不认1-2号机器了
关闭1-2号机器
关闭idc02-kafka-ds-00
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-00bin]$./zkServer.shstop
JMXenabledbydefault
Usingconfig:/usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stoppingzookeeper...STOPPED
|
关闭idc02-kafka-ds-01
|
1
2
3
4
5
|
[hadoop@idc02-kafka-ds-01 bin]$ ./zkServer.sh stop
JMX enabled by default
Using config: /usr/local/webserver/zookeeper-3.4.6/bin/../conf/zoo.cfg
Stopping zookeeper ... STOPPED
|
再查看
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
|
[hadoop@idc02-kafka-ds-03bin]$echomntr|nclocalhost2181
zk_version 3.4.6-1569965,builton02/20/201409:09GMT
zk_avg_latency 0
zk_max_latency 0
zk_min_latency 0
zk_packets_received5
zk_packets_sent4
zk_num_alive_connections 1
zk_outstanding_requests0
zk_server_stateleader
zk_znode_count 4
zk_watch_count 0
zk_ephemerals_count0
zk_approximate_data_size 27
zk_open_file_descriptor_count 27
zk_max_file_descriptor_count 65535
zk_followers 2
zk_synced_followers2
zk_pending_syncs 0
|
没有任何影响
实验成功