When workers fail to launch how do you find out why

[jpfairbanks]

In the example for the yarn cluster manager, I get a timeout when launching the workers. How can I find out why they are failing to launch in order to debug it? I guess i need to look at some log files somewhere.

[tanmaykm]

There may be some errors logged in Yarn logs. In a standalone installation, it is usually the $HADOOP_HOME/logs folder.

While launching a lot of workers, the error are usually:

• Not enough resources. Yarn can not fulfill the request. Yarn would log these errors.
• Default time-out is too less. Specify a higher launch_timeout while creating the cluster manager.