Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Zookeeper, mail # dev - Where are we in ZOOKEEPER-1416


Copy link to this message
-
RE: Where are we in ZOOKEEPER-1416
陈迪豪 2014-01-28, 07:02

Thanks for the detailed explanation.

Now the problem might be settled for me. The HBase requirement I asked should
be discussed in another thread. We watch one znode and wait for notification, but
it would be more "robust" for us to know all the changes at this znode. So if we
really need it, we can do it on the top of ZK.

Thanks again for all your help. I'm the newbie of community but would like to help
develop HBase and ZooKeeper anyway.

________________________________________
From: Thawan Kooburat [[EMAIL PROTECTED]]
Sent: Tuesday, January 28, 2014 1:28 PM
To: [EMAIL PROTECTED]
Subject: Re: Where are we in ZOOKEEPER-1416
I don't understand about how Hbase works but I will try to answer based on
your abstraction here.

As other people already said, if you want to use ZK's notification (watch)
of a single znode as mechanism to drive a state transition of another peer
then it is not going to work.

If want you really need is a reliable notification, ZK doesn't not provide
this mechanism directly. However, you might be able to implement it on top
of ZK by storing each state transition as an individual znode, then it
might be possible to have a peer subscribe to all the state transition.

"Suitable" is a tricky word. You might be able to find other systems which
provide the exact mechanism that you want but then you have to evaluate if
it fit your reliability, throughput or other requirements or not.