Details
-
Improvement
-
Resolution: Fixed
-
Major
-
None
-
5.1.1
Description
There are times when a client query causes the Query service to crash/panic and it is very difficult to identify the offending query. For example, a difficult scenario is when there is production impact due to intermittent crashes of the Query service and we cannot restore Query into a stable state because of the challenge in identifying the offending query in a time-efficient manner.
It would be nice to have a log, similar to active and completed requests, whereby we can log suspect queries that potentially are culprits that caused a crash.
Perhaps this is a simplistic view regarding a possible solution, but we could log every inbound query in a list (similar to active_requests and completed_requests) and remove them when they successfully complete. Queries that generate syntax errors would be ignored. Although this does not exactly identify which query causes a panic/crash, at least we can narrow down the list of queries that could be the culprit and speed up the process of identifying the problematic query.