Duo Zhang commented on HBASE-16835:

Michael Stack
We have to do the watcher? Its a mistake, no? We should do what asynchbase does 
where it just gets what it needs to run from zk then lets go of the connection 
until error and it needs to refresh... then reads again.

Maybe we don't have to make the same mistake here?

All clients having a watcher doesn't scale or at least lags.

Duo Zhang
Using watcher could make the code clean as we can always get a value(maybe 
stale). And I think zk can handle a large set of connections to it?

Duo Zhang
And now only the clusterId is used I think, it does not use watcher internally. 
Let's just remove other stuffs in this class to not prevent the patch to be 
committed. Let's discuss the zookeeper thing later.

Michael Stack

> Revisit the zookeeper usage at client side
> ------------------------------------------
>                 Key: HBASE-16835
>                 URL: https://issues.apache.org/jira/browse/HBASE-16835
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Duo Zhang
> Watcher or not.
> Curator or not.
> Keep connection or not.
> ...

This message was sent by Atlassian JIRA

Reply via email to