On Tue, Dec 17, 2013 at 11:52 PM, Sean Busbey <[email protected]>wrote:
> On Tue, Dec 17, 2013 at 10:44 PM, Keith Turner <[email protected]> wrote: > > > On Tue, Dec 17, 2013 at 11:39 PM, Josh Elser <[email protected]> > wrote: > > > > > > I was not using the correct terminology for what I was thinking, but I am > > not sure what the correct terminology is. I do not want to be limited by > > the memory of one machine for hdfs metadata and I want HA. So would it > be > > correct to say that I want to test with multiple HA nameservices? For > > example test 1.6 w/ three nameservices, where each nameservice has two HA > > namenodes. This would allow hdfs metadata to spread across the memory > of > > three machines. > > > > > These HDFS nameservices are independent HDFS instances using the new > span-HDFS-instances feature in Accumulo 1.6? Or they are nameservices using > HDFS federation across HA pairs? > HDFS federation across HA pairs, describes what I am thinking. I did want the HA pairs to share datanodes as Ravi mentioned. So I would like to test w/ 2 or more federated HA pairs. > > If the former, I think the terminology would be "spanning HDFS clusters." > > If the latter, it's HDFS federation[1]. > > [1]: > > http://hadoop.apache.org/docs/r2.2.0/hadoop-project-dist/hadoop-hdfs/Federation.html > > > -- > Sean >
