HyukjinKwon commented on a change in pull request #30789:
URL: https://github.com/apache/spark/pull/30789#discussion_r543808416
##########
File path: docs/structured-streaming-programming-guide.md
##########
@@ -1689,6 +1689,12 @@ hence the number is not same as the number of original
input rows. You'd like to
There's a known workaround: split your streaming query into multiple queries
per stateful operator, and ensure
end-to-end exactly once per query. Ensuring end-to-end exactly once for the
last query is optional.
+### State Store and task locality
+
+The stateful operations stores states for events in state stores of executors.
State stores occupies resources such as memory and disk space to store the
states. So it is more efficient to keep a state store provider running in the
same executor across different streaming batches. Changing the location of
state store provider requires to load from checkpointed states from HDFS in the
new executor. The stateful operations in Structured Streaming queries rely on
the preferred location feature of Spark's RDD to run the state store provider
on the same executor. However, generally the preferred location is not a hard
requirement and it is still possible that Spark schedules tasks to the
executors other than the preferred ones. In this case, Spark will load state
store providers from checkpointed states on HDFS to new executors. The state
store providers ran in previous batch will not be unloaded immediately. If in
next batch the corresponding state store provider is scheduled on this ex
ecutor again, it could reuse the previous states and save the time of loading
checkpointed state. Spark runs a maintenance task which checks and unloads the
state store providers which are inactive on the executors.
+
+For some use cases like processing very large state data, loading new state
store providers from checkpointed states can be very time-consuming and
inefficient. By changing the Spark configs related to task scheduling, for
example `spark.locality.wait`, users can config Spark how long to wait to
launch data-local task. For stateful operations in Structured Streaming, it can
be used to let state store providers running on the same executors across
batches.
Review comment:
maybe ...
- `For some use cases like processing` -> `For some use cases such as
processing`
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
