github-actions[bot] commented on code in PR #66788:
URL: https://github.com/apache/doris/pull/66788#discussion_r3794852923
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/load/NereidsStreamLoadPlanner.java:
##########
@@ -342,4 +342,11 @@ public TPipelineFragmentParams plan(TUniqueId loadId, int
fragmentInstanceIdInde
params.setEnableTso(destTable.enableTso());
return params;
}
+
+ static TQueryOptions getQueryOptions(NereidsLoadTaskInfo taskInfo) {
+ if (taskInfo instanceof NereidsRoutineLoadTaskInfo) {
Review Comment:
[P2] Keep multi-table Routine Load on the asynchronous defaults
This `instanceof` excludes only the single-table routine representation.
`streamLoadMultiTablePut()` still creates a `NereidsStreamLoadTask` and
overlays its persisted `RoutineLoadJob` via `setMultiTableBaseTaskInfo()`, so
it falls through and copies the FE's current global snapshot. Changing
`enable_hyperscan_fallback` globally can therefore change or fail later tasks
of an existing multi-table routine job, while the equivalent single-table job
keeps the intended fresh `TQueryOptions` default. Please make asynchronous task
provenance explicit and cover both routine representations in the
exclusion/test.
##########
fe/fe-core/src/main/java/org/apache/doris/nereids/load/NereidsStreamLoadPlanner.java:
##########
@@ -342,4 +342,11 @@ public TPipelineFragmentParams plan(TUniqueId loadId, int
fragmentInstanceIdInde
params.setEnableTso(destTable.enableTso());
return params;
}
+
+ static TQueryOptions getQueryOptions(NereidsLoadTaskInfo taskInfo) {
+ if (taskInfo instanceof NereidsRoutineLoadTaskInfo) {
+ return new TQueryOptions();
+ }
+ return ConnectContext.get().getSessionVariable().toThrift();
Review Comment:
[P1] Copy only the intended policy into Stream Load options
This copies every globally settable session option into an RPC Stream Load,
although this planner previously started from a fresh `TQueryOptions`. For
example, `StreamLoadHandler` creates its request context from the global
session defaults, `toThrift()` carries `dry_run_query=true`, and both
tablet-writer `write()` implementations then return OK without writing any
input block. A normal HTTP Stream Load can therefore report success while
ingesting no rows after that unrelated global is enabled;
`truncate_char_or_varchar_columns` similarly starts mutating load data before
strict validation. Please keep fresh load options and copy only
`enable_hyperscan_fallback` from the request snapshot for genuinely synchronous
tasks, with a test proving unrelated globals stay at their load defaults.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
