Hey,

I forward this to hama-user if you don't mind.
If you are using 0.3.0 you should have a look at the 0.4.0 release
candidate. [1]
0.3.0 had some nasty problems with the barrier sync. This should be fixed
in 0.4.0.
There is one thing which is different between the localrunner and the
distributed mode: Serialization.
Your messages must be Writable (from Hadoop Writable) and they need to
properly implement read/write.
This is not used in the localrunner, because it just sends objects between
threads without the serialization.
Also the Hadoop version must be 0.20.2, other versions sadly do not work.

But when we configure the distribute model, we meet some problem.


It would be great if you can be a bit more specific what problem you have
met.

For example, the special configuration of starting hadoop on all distribute
> computer?


In Hadoop there are start scripts under the bin folder. What you need is
"start-dfs.sh". It looks into the "conf/slaves" file and starts daemons on
the mapped server, therefore it needs passwordless ssh. Just like hama.

Could you give me some suggestion about the configuration of hama?


What exactly do you want to hear? :)

Hope that it helps you a bit!

Greetings from Germany,
Thomas


[1] http://people.apache.org/~edwardyoon/dist/0.4-RC5/


2012/2/16 XINSHENG LI <[email protected]>

> Hi thomas:
>
> I am ready to do some research on the hama.
>
> With the help of your blog, I have finished some work with hama on local
> model.
>
> But when we configure the distribute model, we meet some problem.
>
> We are not sure where those problem came from?
>
> Could you give me some suggestion about the configuration of hama?
>
> For example, the special configuration of starting hadoop on all
> distribute computer?  or some special configuration of hama?
>
> Thanks
>
> Xinsheng
>
>
>


-- 
Thomas Jungblut
Berlin <[email protected]>

Reply via email to