Just confirming that following Andrei's trick I was able to run a heterogeneous cluster on EC2.
Thanks for the help, Andrii On Tue, May 15, 2012 at 11:08 PM, Andrii Vozniuk <and...@vozniuk.com> wrote: > Andrei, thanks for the quick response! > > I'll give it a try and reply. What about a general case when I need N > subclusters? > > Cheers, > Andrii > > > On Tue, May 15, 2012 at 10:57 PM, Andrei Savu <savu.and...@gmail.com>wrote: > >> Yes, in trunk. You need the following trick: >> >> whirr.instance-templates=1 hadoop-namenode+hadoop-jobtracker, 5 >> hadoop-datanode+hadoop-tasktracker, 5 >> noop+hadoop-datanode+hadoop-tasktracker >> >> whirr.templates.hadoop-datanode+hadoop.tasktracker.hardware-id=m1.medium >> >> whirr.templates.noop+hadoop-datanode+hadoop.tasktracker.hardware-id=m1.large >> >> Cheers, >> >> -- Andrei Savu / axemblr.com / Tools for Clouds >> >> On Tue, May 15, 2012 at 11:53 PM, Andrii Vozniuk <and...@vozniuk.com> >> wrote: >> >> > Hi all, >> > >> > For research reasons I would like to run Hadoop MapReduce on a >> > heterogeneous EC2 cluster placing 5 datanodes+tasktrackers on m1.medium >> and >> > 5 datanodes+tasktrackers on m1.large instances. >> > >> > I know that in https://issues.apache.org/jira/browse/WHIRR-332 some >> work >> > was done towards using different instances depending on a role. But in >> my >> > case I need to use different instances for the same role. >> > >> > Is this level of customization supported? If not, does the whole >> > architecture supports possible introduction of this feature? >> > >> > -- >> > Best regards >> > Andrii Vozniuk >> > >> > > > > -- > Best regards > Andrii Vozniuk > > > -- Best regards Andrii Vozniuk
