Zookeeper 适合这种场景吗？
这是我的要求。
我在一个集群中有一堆机器（比如说 4 - A、B、C、D）。
A 的工作是轮询数据库以获取更新（因此，A 可能会在订单表中查找新订单）。
一旦 A 收到更新，它就会验证 B、C、D 中哪一个相对空闲（我猜负载平衡是正确的词）。然后它命令 B、C、D 之一开始处理订单。 A 跟踪 B/C/D 上正在处理哪个订单。
B、C、D 就像从属设备 - 它们仅接收来自 A 的更新，并在完成订单处理后通知 A。如果 A 宕机（由于网络问题或其他原因），B、C、D 中的一个将成为主节点并执行 A 的职责。A 关于正在运行的作业的元数据也会定期备份到备份节点 E。 B/C/D 成为新的 master，它会从 E 读取元数据。
我知道这听起来有点像 hadoop，但是订单处理无法适应 MapReduce 模型，所以我正在寻找利用 ZooKeeper 等其他框架的方法这可以帮助协调A、B、C 和 D.
ZooKeeper 适合这里吗？
Here's my requirement.
I have got a bunch of machines in a cluster ( say about 4 - A,B,C,D).
A's job is to poll a database for updates (so, A could be looking for new orders in order table).
Once, A receives an update, it verifies which one of B,C,D is relatively free (load balances is the right word I guess). It then orders one of B,C,D to start processing the order. A keep tracks of which order is being processed on B/C/D.
B,C,D are like slaves - they only receive updates from A and notify A, when they are done with order processing. If A goes down (due to network issue or something), One of B,C,D would become a master and perform duties of A. A's metadata about what jobs are running is also periodically backed up to a backup node E. So when B/C/D becomes the new master, it would read metadata from E.
I know it sounds a bit like hadoop, but the order processing cannot fit into a map reduce model, so I am looking for ways to leveraging other frameworks like ZooKeeper which can help with the coordination between A,B,C and D.
Is ZooKeeper a fit here?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
Zookeeper 是解决协调问题的自然选择。
以下食谱可用于您的用例。
分布式队列可用于任务调度。
看起来像 领导者选举问题
您可以使用 Zookeeper 来存储元数据。
Zookeeper is a natural choice for coordination problems.
Following recipes can be used for your use cases.
Distributed Queue can be used for task scheduling.
Looks like a leader election problem
You can use zookeeper for storing metadata.
Zookeeper 非常适合集群同步（例如主选择）。另一个可以帮助您的相关（Zookeeper 子项目）是 bookkeeper
请注意，hadoop 不使用 Zookeeper（0.23 版本使用了，但尚未发布） - HBase 即使在当前和以前的版本中也使用它
Zookeeper is a good fit for cluster synchronization (e.g. master selection). Another related (sub-project of Zookeeper) that can help you is bookkeeper
Note that hadoop doesn't use zookeeper (version 0.23 does but it isn't released yet) - HBase does use it even in current and previous versions