Describe the bug

When users cancel spark load, it may throw a exception and fail to kill yarn application.

To Reproduce

Steps to reproduce the behavior:

1. Make sure your hadoop file system in yarn cluster needs hadoop ugi
2. Submit a spark load with mysql-client
3. Cancel load
4. Fe throws exception, but the yarn application is still running.

Expected behavior

The application should be killed as soon as user cancel load.

Screenshots