2013/3/25 Eli Reisman <[email protected]> > I have had trouble with the job-local ZK instances myself, > including using the 1.0.x Hadoop clusters.
For running it locally it seems to be more stable to run a separate zookeeper. I always went for doing so, but that is quite simple. If you use a distribution like CDH, just do a yum/apt-get install hadoop-zookeeper-server -y; service hadoop-zookeeper-server start and use that in your job config. If you want to do it by hand, just download the latest production version and run the script to start the server by hand. HTH -- André
