Hi, could a committer take a look at https://github.com/StarRocks/starrocks/pull/78864? It removes an unconditional

hdfsHSync()

on HDFS file close that stalls INSERTs into Hive/Iceberg on busy DataNodes.

@kevincai

already approved; one more approval is needed. Thanks!

hi starrocks reviewers, can you please help review this bugfix PR here github.com/StarRocks/starrocks/pull/79339 🙏

👋 Looking for a feedback on this PR https://starrocks.slack.com/archives/C02FAD0JSSD/p1783529023265349 It's big feature gap in StarRocks compared to Doris and it prevents to use some advanced Flink capabilities without using intermediary kafka topic. Happy to collaborate and looking forward to any additional feedback.

Hey! Please take a look when you have a time: github.com/StarRocks/starrocks/pull/78519

Looking for feedback on these PR : github.com/StarRocks/starrocks/pull/78076 github.com/StarRocks/starrocks/pull/79753

Would really appreciate if someone could take a look at these two PRs, it would unblock us in our auditing effort github.com/StarRocks/starrocks/pull/79871 github.com/StarRocks/fe-plugins-auditloader/pull/32

Hi, Looking for help to review this bugfix PR github.com/StarRocks/starrocks/pull/79875

What's the best way to report a security vulnerability in StarRocks?

Hi, It would be great if someone could please help reviewing the bugfix PR github.com/StarRocks/starrocks/pull/79875. Thanks!

Hey, can someone please chexk the bugfix PR https://github.com/StarRocks/starrocks/pull/79967 TIA

Could someone please take a look at this backport PR ? The original one made by mergify had some issues, so I manually made one github.com/StarRocks/starrocks/pull/79933

Hi, Can someone please review the bugfix PR github.com/StarRocks/starrocks/pull/79875. Thanks!

Can someone please have a look at this PR - https://github.com/StarRocks/starrocks/pull/79967

Can some one pls take a look at this PR: github.com/StarRocks/starrocks/pull/80046

Can we get a review here please, the PR has been around since April and we could really use this fix which is ready to merge? github.com/StarRocks/starrocks/pull/71295

Hey, hey! Can someone please review this PR github.com/StarRocks/starrocks/pull/78519 It is about masking catalog credentials

Hi. Is there a particular process for getting PRs reviewed and merged? There's a couple of PRs open that we're interested in, provided by external contributors, but they don't seem to get any attention to move them on despite reviews being requested and/or messages in this channel. I appreciate there appears to be 900+ open PRs, but if people are actively trying to contribute and providing PRs then it would be nice to see them get some attention as they're usually providing a PR to help with an issue that they are experiencing directly. The contributing guidelines don't really give any information as to how to get PRs moved along. Could someone please advise?

Hi, everyone! I’ve opened a PR to update the description of the Iceberg metadata refresh configuration. The current description only mentions the Hive catalog, but the configuration now applies to various external catalogs, including Iceberg, Delta Lake, Paimon, and others. github.com/StarRocks/starrocks/pull/80215 Please take a look and share your feedback. Thanks!

Hi, everyone! I’d appreciate a review on this pull request when someone has a moment: github.com/StarRocks/starrocks/pull/80115 What: It adds an opt-in session variable,

enable_gather_fragment_child_affinity

(default

false

). When it’s on, the gather/result fragment is placed on a BE already used by its child fragments instead of being picked by global round-robin. Why: In shared-nothing clusters that use

labels.location

for physical isolation, the result fragment of a SELECT can currently land on a BE outside the tenant’s label (github.com/StarRocks/starrocks/issues/79937). Scope: Changes

RemoteFragmentAssignmentStrategy

and

SessionVariable

, with unit tests and docs. Default behavior is unchanged. Thanks!

Hi everyone! Could a maintainer take a look at #76228 (Iceberg V3 deletion vector reads)? github.com/StarRocks/starrocks/pull/76228 We're hitting this in production on 4.1.6 with Databricks UniForm tables: any table with DML fails with "Iceberg V3 Deletion Vectors are not supported." The PR has been open since July and needs a rebase. I left the details in a GitHub comment and I'm happy to help test or take over the rebase. Thanks!

Hi everyone! could I please get some eyes on this: github.com/StarRocks/starrocks/pull/71749 it has been stalled since I addressed the comments in July. 🙂 Thank you so much in advance!!

Hi everyone! Could someone help me get a review or some guidance on github.com/StarRocks/starrocks/pull/78355 (pluggable Group Provider)? It’s been open since August 28 without any review. I’ve kept it up to date with the recent authentication changes, including #80942. The PR addresses a production use case where group membership must be resolved from an external identity service because the authenticated JWT doesn’t contain the required groups. I’d particularly appreciate feedback on whether the proposed extension mechanism is an acceptable direction for StarRocks, or if there’s a preferred alternative. Happy to adjust or split the PR if needed. Thanks!

Hello, guys! I've opened a PR to support

UNSET

for catalog properties in StarRocks. github.com/StarRocks/starrocks/pull/81763 Please take a look at and left some reviews. Thanks!

StarRocks UDAF returns empty result I tried the SumInt UDAF on docs.starrocks.io/docs/sql-reference/sql-functions/JAVA_UDF when I run it, I got empty result even the table has data. see below CREATE TABLE t1 (id INT, m MAPBIGINT,BIGINT); INSERT INTO t1 VALUES (1, map{1:10, 2:20}), (2, map{2:5, 3:30}), (3, map{1:1}); select SUM_INT(id) from t1; +-------------+ | sum_int(id) | +-------------+ | 0 | +-------------+ but if run SET new_planner_agg_stage = 1; first then run the UDF, I got result SET new_planner_agg_stage = 1; select SUM_INT(id) from t1; +-------------+ | sum_int(id) | +-------------+ | 6 | +-------------+ when I looked into the be log , it has the following exception: be.INFO2815216W20261027 230919.114663 128550699832896 java_udf.cpp:379] batch_create_bytebuf,java.lang.NoSuchMethodException: java.nio.DirectByteBuffer.init(long,int) My StarRocks version is 4.0.10-5b36164 and Java version is v21 If I run the same UDF on StarRocks 4.0.10-5b36164 with Java v17, the UDF has the right result. Is there some issues with StarRocks v4.0.10 and Java v21 for UDF?

Hi guys 👋 We are using StarRocks heavily in production and one important thing for us is to be able to use it with Flink 2.3... There is this PR here github.com/StarRocks/starrocks-connector-for-apache-flink/pull/503 that is around since July. I have already reviewed the PR twice and I have also deployed and tested it in our testing enviroments with both Flink 1.20 and 2.3.. Can someone trigger the CI workflow and overall help us merge it please? Thanks in advance 🙏

Hi All can you please reivew PR github.com/StarRocks/starrocks/pull/78705 I'm adding basic support to read SSE-C encrypted iceberg tables

Hi. I made a small contribution to the flink connector so to be able to support credential rotation. Please let me know if you have any questions and if the contribution will be accepted. github.com/StarRocks/starrocks-connector-for-apache-flink/pull/529

Hello, guys! I’ve opened a new PR to fix how Iceberg views are reported. https://github.com/StarRocks/starrocks/pull/82226 Currently,

SHOW FULL TABLES

reports Iceberg views as

BASE TABLE

. This PR fixes that behavior so they are correctly reported as

VIEW

. This change may affect some catalog management behavior, so please keep that in mind when reviewing. Thanks!

Hi All can you please review PRgithub.com/StarRocks/starrocks/pull/82117

Hi all! Could someone familiar with sync MVs take a look at github.com/StarRocks/starrocks/pull/82421 (fixes #82213)? After a normal FE leader change, or a restart before a checkpoint, INSERT into a table with a sync MV built on expressions (e.g.

bitmap_union(to_bitmap(x))

,

hll_union(hll_hash(x))

) fails: the definition replayed from the journal is never analyzed. This started with #49528 and affects 3.5 through 4.2. The fix runs the same analysis as image load when the index is installed from the journal, in restore and in RECOVER. Checked with UT and on a 3-FE cluster (restart and leader stop).