Long time :)
Yeah, my understanding was same as whatever you have written. But today I
noticed that Hbase is picking HADOOP_HOME from bashrc when I was trying to
run it in standalone mode(on my local FS). Although all the Hadoop daemons
were stopped and Hbase was configured to run in standalone mode, it was
still trying to reach "hdfs://localhost:9000" and hence getting :
java.net.ConnectException: Call to localhost/127.0.0.1:9000 failed on
connection exception: .
So I arrived at the conclusion that if I have HADOOP_HOME set in bashrc,
Hbase will try to search for the NN neglecting the properties set in
hbase-site.xml. And once I commented out HADOOP_HOME in bashrc, everything
started behaving just perfectly.
On Sun, May 12, 2013 at 2:02 AM, Stack <[EMAIL PROTECTED]> wrote:
> On Sat, May 11, 2013 at 1:02 PM, Mohammad Tariq <[EMAIL PROTECTED]>
> > Hello Aman,
> > Thank you so much for the quick response. But why would that happen?
> > mean the required env variables are present in hbase-env.sh already. What
> > is the need to source bashrc?
> > Consider a scenario wherein you want to run Hbase in standalone mode. You
> > have a Hadoop setup on the same machine and HADOOP_HOME is set in bashrc,
> > but you don't want to use hadoop for some reason. In that case Hbase will
> > face connection issues as it'll try to contact the Hadoop host(as
> > HADOOP_HOME is present in bashrc) which is not running because it is
> > standalone setup. On the other hand if you are running Hbase in pseudo
> > distributed mode and if you haven't set HADOOP_HOME in bashrc, it would
> > still work.
> > I'm sorry to be a pest of questions. I am actually not able to understand
> > this. Pardon my ignorance.
> Hey Tariq:
> We generally try to avoid picking up anything from the environment. This
> is why you have define where you want to get your JAVA, etc., from in the
> hbase configuration. Is there something in particular that we are finding
> in .bashrc that you notice?