My 2 cents: * Master may or may not run on the same machine as a region server. That depends more on how your distribute your tasks across your cluster than anything * HDFS vs local filesystem: these are not really comparable. If you run in the local file system, you can only run on one node. If you use NFS, or another distributed file system, you have to ensure that you have all of the paths set up correctly so that you don't accidently delete important files (each server should have its own directories for HDFS data and tmp files). * also note that some OS'es delete /tmp on reboot. This could be disconcerting.
Dave -----Original Message----- From: Laurent Hatier [mailto:laurent.hat...@gmail.com] Sent: Wednesday, June 22, 2011 2:14 PM To: user@hbase.apache.org Subject: Re: HBase doesn't work.... Yeap ! I have restart all and reinstall all and it works. :) But i have another questions : - I have seen that Master runs with his RegionServer and the others in the cluster. But it's possible to run Master only and the others RegionServer in the cluster? - Is there a great difference between the HDFS and a folder like /tmp to stock the HBase file (table, etc..) ? 2011/6/22 Jeff Whiting <je...@qualtrics.com> > I'm not sure what your specific problem is but these are the steps I'd take > to get it up and running: > 1. Setup and get hdfs running (verify it by going to > http://hdfs-host:50070/) > 2. Get zookeeper up and running on 1 host > 3. Get hbase up and running in pseudo distributed mode. See: > http://hbase.apache.org/book.**html#distributed<http://hbase.apache.org/book.html#distributed> > . > > Once you have pseudo distributed mode working, all you have to do is add > more nodes and everything works. Make sure dns / hosts file is setup > correctly, if not things wont work. As far as zookeeper is concerned you > would want 1 or 3 zk nodes but not 2. For a cluster your size 1 is plenty. > I'd also update to cdh3u0 as that is the most recent. If you still > can't get pseudo distributed mode working I'd fall back to getting hbase up > and running with just a single node running on the local file system--see > http://hbase.apache.org/book.**html#quickstart<http://hbase.apache.org/book.html#quickstart> > . > > > Good luck, > ~Jeff > > > On 6/22/2011 2:34 AM, Laurent Hatier wrote: > >> And one thing : i have set the hbase-env like this : JAVA_HOME = >> /usr/lib/jvm/java-6-sun, HBASE_MANAGE_ZK = true, HBASE_CLASSPATH = >> /usr/lib/hadoop/conf.... and the hbase-site : hbase.cluster.distributed : >> true and hbase.zookeeper.quorum : IP adresses of my 3 nodes >> >> 2011/6/22 Laurent Hatier<laurent.hatier@gmail.**com<laurent.hat...@gmail.com> >> > >> >> My conf sorry : HBase 0.90.1 with cloudera& Hadoop 0.20 with cloudera >>> (package cdh3b4) >>> >>> >>> 2011/6/22 Laurent >>> Hatier<laurent.hatier@gmail.**com<laurent.hat...@gmail.com> >>> > >>> >>> Hi all, >>>> >>>> I know it's maybe ridiculous but i have problems with HBase installation >>>> in cluster. I will have follow 2-3 tutorials on it but it doesn't work. >>>> Well, first i would like to know if it possible to have 2 servers in >>>> cluster with HBase. I ask this because i have seen that we must have 3 >>>> servers for Zookeeper. >>>> Secondly, i don't have really understood the Zookeeper definition ^^ >>>> (maybe this one can be the first ^^) >>>> I use HDFS for root.dir but there is a lot of difference between HDFS >>>> and >>>> /tmp for example ? >>>> I start "hbase master start" like this and "hbase regionserver start" on >>>> the other node. >>>> I have already insert IP adresses in the regionserver file. >>>> Thirdly, i have question about HDFS : when i start hbase on 1 server it >>>> works (root.dir use HDFS) ; but when i want to see the hadoop system >>>> with >>>> localhost:50030 (and of course i have starting all) it doesn't work... >>>> There >>>> is a relation ? >>>> >>>> If anyone can answer me one of these (or all ^^) thanks to you ! >>>> >>>> >>>> >>>> -- >>>> Laurent HATIER >>>> Étudiant en 2e année du Cycle Ingénieur à l'EISTI >>>> >>>> >>> >>> -- >>> Laurent HATIER >>> Étudiant en 2e année du Cycle Ingénieur à l'EISTI >>> >>> >> >> > -- > Jeff Whiting > Qualtrics Senior Software Engineer > je...@qualtrics.com > > -- Laurent HATIER Étudiant en 2e année du Cycle Ingénieur à l'EISTI
