Hyukjin Kwon created SPARK-61064:
------------------------------------
Summary: Reattach and raise DEADLINE_EXCEEDED when a per-RPC
deadline passes before the response headers in the Rust client as PySpark does
Key: SPARK-61064
URL: https://issues.apache.org/jira/browse/SPARK-61064
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: 5.0.0
Reporter: Hyukjin Kwon
The Rust client applies PySpark's per-RPC deadlines with tonic's grpc-timeout,
which tonic enforces only until the response headers and reports as CANCELLED
"Timeout expired". So when a deadline passes before the headers, a reattachable
execution fails instead of reattaching as PySpark does, and a unary RPC raises
CANCELLED instead of DEADLINE_EXCEEDED with PySpark's message. The client also
does not execute again when a reattach finds INVALID_HANDLE.OPERATION_NOT_FOUND
before any response, as PySpark does.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
