Description
We used azkaban to submit many yarn jobs, and the /tmp directory will have many hadoop-unjar directories. Sometimes the hadoop-unjar directory on the azkaban machine takes up a lot of space, but we do not know which process generated this directory. In order to solve this problem, we add the hadoop process id to the suffix of the hadoop-unjar directory.
- hadoop process id
10554 org.apache.hadoop.util.RunJar
- hadoop-unjar directory name
hadoop-unjar8020753511094521686-10554