Andrew Kyle Purtell created PHOENIX-8050:
--------------------------------------------
Summary: Highly selective queries can severely underreport scan
metrics
Key: PHOENIX-8050
URL: https://issues.apache.org/jira/browse/PHOENIX-8050
Project: Phoenix
Issue Type: Bug
Affects Versions: 5.3.2
Reporter: Andrew Kyle Purtell
Assignee: Andrew Kyle Purtell
Fix For: 5.4.0, 5.3.3
Top-N and offset scanners pre-read rows at initialization, recording their scan
metrics in a local context intended for HBase's RPC response. {{next()}} checks
{{isFilterDone()}} before merging this local context into the active RPC
context, so if all rows in a region are filtered out during scanner setup,
{{next()}} exits immediately, discarding those initialization metrics before
they can be recorded.
As a result, queries that filter out entire regions severely underreport scan
metrics such as {{ROWS_SCANNED}} and {{{}ROWS_FILTERED{}}}.
The fix flushes the initialization metrics into the RPC context before checking
{{{}isFilterDone(){}}}. This ensures scan metrics accounting is accurate across
empty regions while preserving early-exit behavior and query results.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
