Hyukjin Kwon created SPARK-61187:
------------------------------------
Summary: Raise NO_ACTIVE_SESSION when a stream of a stopped
session reattaches, as PySpark does
Key: SPARK-61187
URL: https://issues.apache.org/jira/browse/SPARK-61187
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
After stop() closes the client, the RPCs of the session raise
NO_ACTIVE_SESSION, but a result stream that was still open at stop() kept
reading: when the server ended the stream (for example after
senderMaxStreamDuration) and the client had to reattach, ReattachExecute went
out and the rows kept coming. The reattach and an ExecutePlan sent again
bypassed the closed-client check. PySpark's iterator reattaches through the
closed client's stub, which raises NO_ACTIVE_SESSION, so the read fails with
that error. The fix checks the closed state before each reattach attempt and
ends the stream with that error.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
