@arina-ielchiieva, very good point. Having that information is very useful regardless of the source of error. How could we provide that generically without the need to wrap each bit of code in a "in case something goes wrong" try/catch block? For example, suppose we get an NPE or OOM. Should we catch all those?
Could each operator (including scanners) provide a context which can be dumped to the log along with the stack trace? Often, we try to recreate the fragment operators from the stack trace. Would be better to see a stack of operators, each with it's critical info. The fragment exec walks the operator tree to shut down each operator. I could also do so to create the operator stack with related info. Is this something we could consider? Would have been very helpful the many times I tried to track down a failure found in a QA test or user case. [ Full content available at: https://github.com/apache/drill/pull/1455 ] This message was relayed via gitbox.apache.org for [email protected]
