Hi Terry, I believe dfs is usually referring to local file system, but they're mostly just aliases under storage-plugins.json that refers to the actual storage setting.
You can refer here for more information how to configure them: https://cwiki.apache.org/confluence/display/DRILL/Connecting+Apache+Drill+to+Data+Sources And yes for zk hosts, it's host:port seperated by commas, ie: localhost:2181,localhost:2191 Tim On Mon, Jun 16, 2014 at 2:00 PM, Terry Healy <[email protected]> wrote: > Thanks Tim. > > I'm starting with 6 nodes. Got as far as trying to registering file systems, > but when I start Drill, all the storage plugins fail. > > I'm most interested to get HDFS up; is that 'dfs' or 'hadoop_simple'? > > I have hBase running on the cluster, but I see this error: > > DEBUG o.a.d.e.s.h.HBaseStoragePluginConfig - Initializing HBase > StoragePlugin configuration with zookeeper quorum 'localhost', port '2181'. > > So I assume I have to set it to my 3 active Zookeepers, in the config here: > > "hbase.zookeeper.quorum" : "localhost", > > So just hostnames separated by commas? > > > > Thanks for your help. > > -Terry > > > On 06/16/2014 04:04 PM, Timothy Chen wrote: >> >> Hi Terry, >> >> You can start testing even with one node if you like, but ofcourse to >> gain more power for parallelizing you want more nodes which is what >> Drill is designed for. >> >> Drill can pull from HDFS from the configured HDFS information as part >> of the storage engine config. More info about that should be available >> in our wiki. >> >> Tim >> >> >> >> On Mon, Jun 16, 2014 at 12:41 PM, Terry Healy <[email protected]> wrote: >>> >>> Hi- >>> >>> Is it necessary to install Drill on all the nodes of an HDFS cluster, or >>> is >>> it OK to start testing with a few? >>> >>> Thanks >>> >
