James Taylor created PHOENIX-4593:

             Summary: Detect and fail queries that are deemed too expensive
                 Key: PHOENIX-4593
                 URL: https://issues.apache.org/jira/browse/PHOENIX-4593
             Project: Phoenix
          Issue Type: Improvement
            Reporter: James Taylor

Based on a conversation over on PHOENIX-1556, we should have configurable 
limits for various query operators:
- max size of client-side order by
- max size of server-side order by
- max size of client-side aggregation
- max size of server-side aggregation
- max bytes processed for an UPSERT SELECT
- max rows deleted by DELETE

Some of these are controlled by the max amount of memory allowed, but this is 
suboptimal as you end up using cluster resources and then failing at runtime. 
Ideally, if we had histograms available (PHOENIX-1178), we could detect at 
compile time if we think the limits will be reached and then disallow them.

This message was sent by Atlassian JIRA

Reply via email to