[TAJO-1625] Recap error propagation system

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.11.0
• Assignee: Hyunsik Choi
• Reporter: Hyunsik Choi
• Created: 27/May/15 08:30
• Updated: 04/Sep/15 02:13
• Resolved: 04/Sep/15 02:13

Description

In the current implementation, our error propagation system is not consistent.
The primary problem comes with RPC. RPC just throws exception with the error message. So it makes error handling and the failure tests very difficult. Another problem is that the exact error status is not conveyed to users.

The solution is to refactor all exceptions used in Tajo to have error code number and stack trace throughout all codes and rpc responses.

I'm surveying all error cases used in Tajo at https://cwiki.apache.org/confluence/display/TAJO/Error+Code. I would appreciate your effort If you help to this survey.

Issue Links

• is related to: TAJO-1699 Tajo Java Client version 2

Sub-Tasks

• Refactor client errors and exceptions: Resolved
• Refactor visitors and planner to throw TajoException: Resolved
• Refine client APIs to throw specific exceptions: Resolved
• Refine JDBC exceptions to better handle exceptional cases.: Resolved
• Some TajoRuntimeException should be restored as TajoException in client side: Resolved
• Remove unused and legacy exceptions: Resolved