hadoop-hdfs-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Vinayakumar B <vinayakum...@apache.org>
Subject Re: Why DFSClient in HA-cluster tries to resolve HA cluster name?
Date Tue, 31 Mar 2015 10:10:44 GMT
Thats a good find Dmitry Simonov.
Even in DFSAdmin, everywhere its used just to print to console.

You can raise a jira and post a patch to refactor this.

-Vinay

On Tue, Mar 31, 2015 at 12:34 PM, Dmitry Simonov <dimmoborgir@gmail.com>
wrote:

> Hello!
>
> Why DFSClient in HA-cluster tries to resolve HA cluster name?
>
> I have just been investigating a problem - FileSystem.get() method
> sometimes takes too long time to complete, about 2-3 seconds, and found
> that:
>
> 1. DFSClient makes call to NameNodeProxies.createProxy()
> 2. createProxy() method, in the "HA case", tries to resolve nameNodeUri to
> IP-address (even if it is a logical nameservice name). It wastes time.
> 3. In the returned object (of ProxyAndInfo<T> class), IP-address is used
> only in DFSAdmin tool (via call to "InetSocketAddress getAddress()"
> method).
>
> Best regards, Dmitrii Simonov.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message