DB Tsai created SPARK-60966:
-------------------------------
Summary: Cover reattachable execution cleanup over gRPC in Spark
Connect Rust
Key: SPARK-60966
URL: https://issues.apache.org/jira/browse/SPARK-60966
Project: Spark
Issue Type: Test
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: DB Tsai
Fix For: connect-rust-4.3.0
SPARK-60739 fixed unreleased executions in the Rust client's reattachable
execution. Add regression coverage that exercises the actual gRPC boundary,
including ReleaseExecute request contents and metadata, without requiring a
live Spark server.
Use a loopback tonic service to cover:
* ResultComplete returns without waiting for the cleanup reply and schedules
one ReleaseAll even when the stream is subsequently dropped.
* Incremental ReleaseUntil and early-drop ReleaseAll, both inside and outside a
Tokio runtime.
* Terminal query errors remain the reported error when cleanup fails.
* Transient cleanup failures retry with the same request and metadata.
* Permanent cleanup failures preserve a successful query result.
* Graceful EOF and transient disconnection reattach before releasing the
execution.
Keep the service fixture in integration tests and add tonic-prost only as a
dev-dependency. This is a test-only follow-up; the production cleanup fix is
already present in master.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
