Another way is to use Java remove debugging feature, which allows you to keep your IDE and source code locally and run a Hadoop daemon with a couple extra JVM flags to bring up debug server within the process.
On Wed, Jan 26, 2011 at 09:22, Steve Loughran <ste...@apache.org> wrote: > On 23/01/11 06:31, Zinab Ahmed Mahmoud Elgendy wrote: >> >> I want to make debugging step by step using eclipse for example. I thought >> about remote debugging but i couldn't use it to debug the slave code. >> > > on your local (unix) desktop set X up to allow anyone: > Â xhost + > > on the target machine install your chosen IDE , but set the DISPLAY > environment variable to your desktop's (hostname, display) > > export DISPLAY=desktop:0.0 > > where desktop should be changed to your desktop's hostname > > bring up eclipse, attach it to whatever process you want. > > The hard part is the TT will spawn new JVMs for new jobs, making binding > harder. > > Personally I'd recommend putting effort into logging and testing, as they > have longer term benefits and scale well. But it is still nice to know what > goes on in worker nodes from time to time >
