Neil Ramaswamy created SPARK-50584:
--------------------------------------
Summary: Modify state eviction metrics in TransformWithState to
represent physical deletions
Key: SPARK-50584
URL: https://issues.apache.org/jira/browse/SPARK-50584
Project: Spark
Issue Type: Improvement
Components: Structured Streaming
Affects Versions: 4.0.0
Reporter: Neil Ramaswamy
In SPARK-50302, we modified the indexing logic for TWS variables with TTL. As
part of these changes, we made the `numRowsRemoved` parameter represent the
number of logical state entries that were evicted as a result of TTL
expiration. However, the `numRowsRemoved` parameter, by its name, should be
about the number of underlying state store _rows_ that were removed. This is a
physical metric, not a logical metric.
The logical metric that represents TTL expiration is
`numValuesRemovedDueToTTLExpiry`, and that can be used for non-physical
metrics. This fix should be made across all state variables—value, map, and
list.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
