hama-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Thomas Jungblut <thomas.jungb...@googlemail.com>
Subject Re: Unable to run MaxFindVertex
Date Wed, 30 May 2012 15:06:15 GMT
Hi Apurv,

frankly it works for me:

part-0000 contains:
D 6
A 6
part-0001 contains:
B 6
and part-0002 contains:
C 6



2012/5/30 Apurv Verma <dapurv5@gmail.com>

> Here is how I think my algo works. Please tell me where I am wrong.
>
>
> @Override
>    public void compute(Iterator<IntWritable> messages) throws IOException {
>      IntWritable currentMax = this.getValue();
>
>
>      System.err.println();
>      System.err.println(getSuperstepCount());
>      System.err.println("At node "+this.getVertexID());
>
> IN THE 0th SUPERSTEP, ALL VERTICES SEND THEIR VALUES TO ALL NEIGHBORING
> VERTICES AND RETURN.
>
>      if(getSuperstepCount() == 0L){
>        System.err.println("sending msg to all neighbors
> ..."+currentMax.toString());
>        sendMessageToNeighbors(currentMax);
>        return;
>      }
>
> //FOR ALL SUBSEQUENT SUPERSTEPS, EACH VERTEX LOOKS AT ALL VALUES IT HAS
> RECEIEVED AND STORES THE MAX VALUE RECEIVED IN currentMax.
>
>      while(messages.hasNext()){
>        IntWritable msg = messages.next();
>        currentMax = (msg.compareTo(currentMax) > 0)?msg:currentMax;
>      }
>
> //IF THERE IS AN UPDATE IN THE CURRENT VALUE OF THE VERTEX, THAT MEANS THAT
> THE VERTEX HAS NOT VOTED TO HALT SO IT SHOULD SEND MESSAGES TO ALL ITS
> NEIGHBORING NODES.
>
>      if(currentMax.compareTo(getValue()) > 0){
>        System.err.println("updating value to ..."+currentMax.toString());
>        this.setValue(currentMax);
>        sendMessageToNeighbors(getValue());
>      }
>      else{
>       IF NOT IT SHOULD SEND NO MESSAGES. i.e. THE VERTEX HAS VOTED TO HALT.
>        System.err.println("voting to halt...");
>      }
>
>      System.err.println();
>    }
>  }
>
> AT THE END I EXPECT TO SEE AN OUTPUT FILE WHICH SHOULD BE SOMETHING LIKE
> THIS
> A 6
> B 6
> C 6
> D 6
>
> But the job fails. And there are 3 output files created part-00000 which is
> empty, part-00001, again empty. and part-00002 which has C 2 written in it.
>
> --
> thanks and regards,
>
> Apurv Verma
> India
>
>
>
>
>
> On Wed, May 30, 2012 at 12:41 PM, Apurv Verma <dapurv5@gmail.com> wrote:
>
> > Yes, I ran that example by copying it verbatim, it ran. Don't know why my
> > MaxFindVertex is not working.
> >
> >
> > --
> > thanks and regards,
> >
> > Apurv Verma
> > India
> >
> >
> >
> >
> >
> > On Wed, May 30, 2012 at 11:31 AM, Thomas Jungblut <
> > thomas.jungblut@googlemail.com> wrote:
> >
> >> Oops, I have read "with aggregator" and not without. I'm sorry Apurv.
> >> Do you had a look at the mindist search/connected component example in
> >> example package? It does actually the same but inverted.
> >>
> >>
> >> 2012/5/30 Apurv Verma <dapurv5@gmail.com>
> >>
> >> > Isn't an aggregator optional?
> >> >
> >> > --
> >> > thanks and regards,
> >> >
> >> > Apurv Verma
> >> > India
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > On Wed, May 30, 2012 at 11:26 AM, Thomas Jungblut <
> >> > thomas.jungblut@googlemail.com> wrote:
> >> >
> >> > > Hi Apurv,
> >> > >
> >> > > there is no aggregator defined in your code.
> >> > >
> >> > > 2012/5/30 Apurv Verma <dapurv5@gmail.com>
> >> > >
> >> > > > Hello,
> >> > > >  After several attempts I was unable to run the MaxFindVertex
> >> without
> >> > > using
> >> > > > Aggregators.
> >> > > >
> >> > > > Here is the source code. http://pastebin.com/VMTXPtRR. I have
> >> changed
> >> > > the
> >> > > > num of bsp tasks to 1,2 and 3 but still it doesn't work. The
input
> >> > graph
> >> > > > file is very small.
> >> > > > Here is the graph file. http://pastebin.com/uAJASSRi
> >> > > >
> >> > > > Here are the logs.
> >> > > >
> >> > > > These are the task logs.
> >> > > >
> >> > > >   1. http://pastebin.com/imMKB6Hs
> >> > > >   2. http://pastebin.com/xvMZris0
> >> > > >   3. http://pastebin.com/PkgMrsdx
> >> > > >
> >> > > >
> >> > > > The bsp master log http://pastebin.com/0uQYP03B
> >> > > >
> >> > > >
> >> > > > The groom log  http://pastebin.com/GYY9sUEv
> >> > > >
> >> > > > Can you please let me know my mistake.
> >> > > >
> >> > > > --
> >> > > > thanks and regards,
> >> > > >
> >> > > > Apurv Verma
> >> > > > India
> >> > > >
> >> > >
> >> > >
> >> > >
> >> > > --
> >> > > Thomas Jungblut
> >> > > Berlin <thomas.jungblut@gmail.com>
> >> > >
> >> >
> >>
> >>
> >>
> >> --
> >> Thomas Jungblut
> >> Berlin <thomas.jungblut@gmail.com>
> >>
> >
> >
>



-- 
Thomas Jungblut
Berlin <thomas.jungblut@gmail.com>

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