[FLINK-12781] REST handler should return full stack trace instead of just error msg

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.7.2
• Fix Version/s: 1.9.0
• Component/s: Runtime / REST
• Labels: pull-request-available
• Assignee: Steven Zhen Wu
• Reporter: Steven Zhen Wu
• Created: 07/Jun/19 21:43
• Updated: 02/Feb/20 21:54
• Resolved: 02/Feb/20 21:54

Description

We use REST api to start a job in Flink cluster.
https://ci.apache.org/projects/flink/flink-docs-stable/monitoring/rest_api.html#jars-jarid-run

When there is exception during job construction, the response payload doesn't contain the full stack trace. 

{"errors":["org.apache.flink.client.program.ProgramInvocationException: The main method caused an error."]}

This problem becomes more serious after FLINK-11134 got released in 1.7.2, because stack trace is completely lost now. FLINK-11134 is doing the right thing. We just need the response payload to contain the full stack trace, which seems to have been an issue/fix since 1.5. we know 1.4 doesn't have this issue and correctly return full stack trace

on the jobmanager log, we only get

2019-06-07 17:42:40,136 ERROR org.apache.flink.runtime.webmonitor.handlers.JarRunHandler    - Exception occurred in REST handler: org.apache.flink.client.program.ProgramInvocationException: The main method caused an error.

Issue Links

• links to: GitHub Pull Request #8665