ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nate Cole <>
Subject Re: Dev workflow
Date Fri, 26 Jun 2015 13:18:15 GMT
Other’s have commented; I personally have development working locally with Eclipse, but it
is a bit more work to get going.  The benefit is you don’t have to build rpm or even jar
- just run locally.  Debugging becomes inline and can recompile while the server is running.

* Install Vagrant only for use with Ambari’s agents.  This is because the agent code is
pretty specific to directory locations and linux commands, etc.
* Run Ambari Server locally out of an IDE
* * This requires setting up your local machine with Postgres
* * Setting several values in to reference local directories instead of
deployed locations.  I have attached the one I use, replace SRC_HOME and ETC_HOME appropriately.
 (May have some deprecated properties, I don’t update it often).
* * Setting database values in the properties file to reference local filesystem.
* * When using an IDE, when you start Ambari, add a directory to the classpath that has the
ETC_HOME directory.
* Have the ability to copy agent python files to vagrant.  I have several helper scripts to
do this with mapped directories.
* Successfully build ambari-web locally.  UI folks use brunch to ease development, for backend
it’s enough to build maven.
* Successfully build ambari-views.
* When the UI is loading and installing a clusters, agents should registered manually, not
bootstrapped.  (openssl directory assumptions don’t match up for me and it was just painful)

I’m not sure if this has ever been documented.  If you choose to go this route and don’t
find any documentation we can get a wiki started for “local development.”  To my knowledge
only a few have done it this way.  You can PM me at<>
to work through any issues you see or setup a hangout to discuss.


  • Unnamed multipart/mixed (inline, None, 0 bytes)
View raw message