stack commented on HBASE-16835:

Thanks for doing this writeup [~Apache9]

Agree that getClusterId can be a get blocking call.

getCurrentNrHRS doesn't seem to be used by anything by a test that is verifying 
'round robinness'. Probably better if we asked the master for this stat. 
Doesn't seem important, used much. Could this be blocking too? 
getMasterInfoPort is different from the port you get from getMasterAddress. The 
former is the Web UI port. The latter is the listening port. Could we add the 
web ui port and other vitals to the master znode data. It is pb so extensible.

Agree the important ones are getMasterAddress and getMetaRegionLocation. The 
latter will change later when we can split meta table but it is fine for now.

Would be cool if we could do away w/ watcher on client.

You think then that we'd have a new Registry API for async use? Could the new 
API underpin the old API?

> 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