I am not entirely sure if that was the intended configuration for the scripts, but that is probably how it currently is since the most common configuration involves the same SPARK_HOME on all machines.
TD On Thu, Feb 13, 2014 at 1:53 PM, Guanhua Yan <gh...@lanl.gov> wrote: > Thanks, TD. It seems that in order to use the magic start-*.sh to launch a > cluster, all the nodes should have the same SPARK_HOME setting. > > Best regards, > - Guanhua > > From: Tathagata Das <tathagata.das1...@gmail.com> > Reply-To: <user@spark.incubator.apache.org> > Date: Thu, 13 Feb 2014 13:12:21 -0800 > To: <user@spark.incubator.apache.org> > Subject: Re: Cluster launch > > You could use sbin/start-slave.sh on the slave machine to launch the > slave. That should use the local SPARK_HOME on the slave machine to launch > the worker correctly. > > TD > > > On Thu, Feb 13, 2014 at 1:09 PM, Guanhua Yan <gh...@lanl.gov> wrote: > >> Hi all: >> >> I was trying to run sbin/start-master.sh and sbin/start-slaves.sh for >> launching a standalone cluster, which contains a linux workstation and a >> mac desktop. On these two computers, the SPARK_HOME directories point to >> different places. When running ./sbin/start-slaves.sh, I got an error >> saying that on the slave machine, the spark directory doesn't exist. I >> guess that in the start-slaves.sh script, the SPARK_HOME configuration on >> the master machine was used when launching spark on the slave machine. >> >> Any clues about how to fix this? >> >> Thank you, >> - Guanhua >> >> >