Hi Lars, I am able to run pseudo-distributed mode from a dev tree. Here is the wiki: https://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-common/SingleCluster.html#Pseudo-Distributed_Operation .
Check out my script pseudo_dist <https://github.com/jzhuge/hadoop-sanity-tests/blob/master/bin/pseudo_dist> to start/stop a pseudo-distributed cluster. Here are the steps: 1. mvn install -DskipTests -DskipShade -Dmaven.javadoc.skip -Pdist -Dtar 2. pseudo_dist start ~/hadoop-sanity-tests/config/insecure/ 3. test_env hdfs dfs -ls /tmp Thanks, On Wed, Jul 19, 2017 at 11:49 PM, Lars Francke <lars.fran...@gmail.com> wrote: > I've already asked in <https://issues.apache.org/jira/browse/HDFS-11596> > but haven't gotten a reply so far so I thought I'd bump it here. > > The issue replaces the compile time dependency of the various HDFS projects > to hdfs-client with a "provided" dependency. > > Unfortunately that means that HDFS cannot be run anymore from source as is > documented in the Wiki (< > https://wiki.apache.org/hadoop/HowToSetupYourDevelopmentEnvironment>) and > as used to be possible before the patch. This is because the hdfs client > classes (e.g. ClientProtocol is the first one that HDFS complains about > during startup) are not in the classpath anymore. > > I wonder how all of you are running Hadoop these days from source? I've > always followed the Wiki instructions but maybe they are out of date and > there's a better way? > > Thanks, > Lars > -- John