DB Tsai created SPARK-60989:
-------------------------------
Summary: Populate configured user context on typed requests in
Spark Connect Rust
Key: SPARK-60989
URL: https://issues.apache.org/jira/browse/SPARK-60989
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: DB Tsai
Fix For: connect-rust-4.3.0
The Spark Connect Rust client can have a user ID configured through
ChannelBuilder, but typed ExecutePlan and AnalyzePlan requests, including the
initial request for reattachable execution, do not populate a missing or empty
UserContext.user_id from that configuration. The configured identity can reach
gRPC metadata while remaining absent from the protobuf request body.
To reproduce, configure a user ID on the channel, issue one of these typed
requests with no UserContext or an empty user_id, and inspect the request
received by a gRPC endpoint. The body should contain the configured user ID
when the caller has not provided one.
The fix should:
* Populate missing user IDs on the affected typed execution and analysis
requests before constructing the reattachable response stream.
* Preserve explicit caller-provided IDs, user names, and context extensions.
* Carry the effective execution context through ReattachExecute and
ReleaseExecute, including cleanup after completion or an early stream drop.
* Leave raw serialized passthrough requests byte-exact, including unknown
protobuf fields.
Artifact attribution is already handled on master by PR #164 through the shared
upload_artifact_requests path; that behavior should be retained when rebasing.
Broader client_type propagation is separate work.
Regression coverage should inspect actual protobuf requests at an in-memory
gRPC endpoint for absent, empty, and explicit contexts; reattach and release
identity; asynchronous cleanup; and unchanged raw passthrough bytes. Existing
regression results on the proposal branch need to be refreshed after rebasing
onto the current execution lifecycle.
Proposed fix and regression coverage by WindsorWZZ:
https://github.com/apache/spark-connect-rust/pull/140
Related artifact attribution change:
https://github.com/apache/spark-connect-rust/pull/164
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
