You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hama.apache.org by Anveshi Charuvaka <an...@gmail.com> on 2013/11/08 03:12:59 UTC

Hama seemingly not using multiple nodes on cluster

Hi All,	
	
	I am trying to run a small test job using HAMA graph api on a cluster configured with 10 nodes. My job creates 9 vertices, but it seems that HAMA is loading all the vertices into a single physical machine instead of distributing them onto different machine. 

	I have attached the program. I simply reads the vertices from the input file, in the format ( <ID> <value> <neighbor_id>), the values and neighbors have no significance in this test program. In the compute method, the vertex finds it's peer name using  
	this.getPeer().getPeerName();
and sets it as the value of the vertex, which is then dumped into the output

	Please help. I would like to make it run, so that the load is taken up by multiple physical machines.





Thanks
Anveshi

Re: Hama seemingly not using multiple nodes on cluster

Posted by Anveshi Charuvaka <an...@gmail.com>.

Thanks Edward for the prompt reply, setting setNumBspTask( ), worked :).

Anveshi


On Nov 7, 2013, at 11:27 PM, Edward J. Yoon wrote:

> Hi,
> 
> Please check whether hama cluster is correctly setup. If cluster is
> correctly setup as a fully distributed mode, you'll see the logs like
> this:
> 
> $ tail -f logs/hama-edward-bspmaster-master.log
> ....
> 2013-11-08 13:21:43,667 INFO org.apache.hama.bsp.BSPMaster: Starting RUNNING
> 2013-11-08 13:21:51,306 INFO org.apache.hama.bsp.BSPMaster:
> groomd_slave1_50000 is added.
> 2013-11-08 13:21:51,317 INFO org.apache.hama.bsp.BSPMaster:
> groomd_slave2_50000 is added.
> ....
> 
> And please set the number of tasks in job configuration.
> 
> gJob.setNumBspTask(3);
> 
> Then it will work. :)
> 
> On Fri, Nov 8, 2013 at 11:12 AM, Anveshi Charuvaka
> <an...@gmail.com> wrote:
>> Hi All,
>> I am trying to run a small test job using HAMA graph api on a cluster
>> configured with 10 nodes. My job creates 9 vertices, but it seems that HAMA
>> is loading all the vertices into a single physical machine instead of
>> distributing them onto different machine.
>> 
>> I have attached the program. I simply reads the vertices from the input
>> file, in the format ( <ID> <value> <neighbor_id>), the values and neighbors
>> have no significance in this test program. In the compute method, the vertex
>> finds it's peer name using
>> this.getPeer().getPeerName();
>> and sets it as the value of the vertex, which is then dumped into the output
>> 
>> Please help. I would like to make it run, so that the load is taken up by
>> multiple physical machines.
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> Thanks
>> Anveshi
>> 
> 
> 
> 
> -- 
> Best Regards, Edward J. Yoon
> @eddieyoon

Re: Hama seemingly not using multiple nodes on cluster

Posted by "Edward J. Yoon" <ed...@apache.org>.

Hi,

Please check whether hama cluster is correctly setup. If cluster is
correctly setup as a fully distributed mode, you'll see the logs like
this:

$ tail -f logs/hama-edward-bspmaster-master.log
....
2013-11-08 13:21:43,667 INFO org.apache.hama.bsp.BSPMaster: Starting RUNNING
2013-11-08 13:21:51,306 INFO org.apache.hama.bsp.BSPMaster:
groomd_slave1_50000 is added.
2013-11-08 13:21:51,317 INFO org.apache.hama.bsp.BSPMaster:
groomd_slave2_50000 is added.
....

And please set the number of tasks in job configuration.

gJob.setNumBspTask(3);

Then it will work. :)

On Fri, Nov 8, 2013 at 11:12 AM, Anveshi Charuvaka
<an...@gmail.com> wrote:
> Hi All,
> I am trying to run a small test job using HAMA graph api on a cluster
> configured with 10 nodes. My job creates 9 vertices, but it seems that HAMA
> is loading all the vertices into a single physical machine instead of
> distributing them onto different machine.
>
> I have attached the program. I simply reads the vertices from the input
> file, in the format ( <ID> <value> <neighbor_id>), the values and neighbors
> have no significance in this test program. In the compute method, the vertex
> finds it's peer name using
> this.getPeer().getPeerName();
> and sets it as the value of the vertex, which is then dumped into the output
>
> Please help. I would like to make it run, so that the load is taken up by
> multiple physical machines.
>
>
>
>
>
>
>
> Thanks
> Anveshi
>



-- 
Best Regards, Edward J. Yoon
@eddieyoon