github-actions[bot] commented on code in PR #69298:
URL: https://github.com/apache/doris/pull/69298#discussion_r4321437263
##########
fe/fe-connector/fe-connector-iceberg/src/main/java/org/apache/doris/connector/iceberg/IcebergScanPlanProvider.java:
##########
@@ -2858,37 +2872,80 @@ private CloseableIterable<FileScanTask>
cacheBackedFileScanTasks(TableScan scan,
if (!ManifestEvaluator.forPartitionFilter(filterExpr, spec,
caseSensitive).eval(manifest)) {
continue;
}
- deleteFiles.addAll(manifestCacheGet(manifest, table, specsById,
statsQueryId).getDeleteFiles());
+ deleteFiles.addAll(manifestCacheGet(manifest, table,
statsQueryId).forScan(spec).getDeleteFiles());
Review Comment:
[P2] Apply the query read budget to delete manifests too
Phase 1 loads delete manifests before the statement-wide semaphore is
obtained, so these reads never acquire it. With
`iceberg_manifest_load_concurrency=1`, `split_batch_size=1`, and batch-mode
`UNION ALL` scans of two v2 tables, the first producer can publish its first
split and start its next cold data-manifest read while the second scan
initializes and reads a cold delete manifest here. Position-only deletes have
`total-equality-deletes=0`, so the earlier equality-field discovery does not
warm that entry. Two reads can therefore overlap despite the new per-query
maximum of one, defeating the sequential setting. Obtain the shared budget
before Phase 1 and apply it to actual delete-manifest loads as well.
--
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]
