[kepler-users] How to verify the DistributedCompositeActor?

[Jianwu Wang]

Hi Daren,

    Please see my reply below.

On 3/6/12 8:38 AM, Daren Thomas wrote:
> I just created a simple workflow (see attachment) that counts up to
> 100 using the DistributedCompositeActore. The composite contains a
> simple +1 to the input port and i use a sequence delay on the local
> (master) side to seed the loop.
>
> This works. But it isn't being distributed. I know, because it works
> even if I remove the network connection - *so*, everything is
> happening locally. I tried removing "localhost" from the list of
> available slaves, but this doesn't help.
   The reason that your workflow doesn't distribute is that there is no 
director inside of the DistributedCompositeActor. If there is no 
director, the actor will behavior as a normal composite actor. Please 
put a director in your DistributedCompositeActor and try it again.
> Also, my other pc doesn't
> show up in the list - I have to add it manually. I know I'm doing
> something wrong here, but what?
>
> Where do I register slaves? The master-slave pdf mentions a registry
> service, but how do I set it up? Also, how can I ensure that a
> component is run distributed?

      By default, each slave will not register itself. If you want to 
enable slave registration, you need to change its configuration at 
master-slave/resources/configurations/configuration.xml. Please change 
the value of item 'registration->register' from 'false' to be 'true'.

      If you have director in the DistributedCompositeActor, it will 
ensure the actor will run distributed.

>
> Oh, and when we're done with that - can I have multiple slaves on the
> same host? This would be great, since it would allow me to run a cpu
> bound task on each of my cores...
     Since DistributedCompositeActor is based on Java RMI and each 
machine can only have one hostname, I don't think you run multiple 
slaves on the same host.
> Regards,
> Daren Thomas
>
>
> _______________________________________________
> Kepler-users mailing list
> Kepler-users at kepler-project.org
> http://lists.nceas.ucsb.edu/kepler/mailman/listinfo/kepler-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nceas.ucsb.edu/kepler/pipermail/kepler-users/attachments/20120307/dc3fc54f/attachment.html>