http://starrocks.io logo
Join Slack
Powered by
<@U0A71G31CDV> when we can expect the release dat...
# questions-and-troubleshooting
s
@Rocky when we can expect the release date for 4.2
r
Based on StarRocks' official release policy and recent version history, here is the expected timeline for version 4.2: * Expected Release Window: October 2026. * Release Cycle: StarRocks typically follows a four-month cycle for minor version releases (e.g., 4.0 to 4.1 to 4.2). * Historical Context: * Version 4.1 was released in June 2026 (specifically June 7, 2026). * Version 4.0 was released in October 2025. * Current Status: As of September 2026, the current stable line is 4.1 (with the latest patch being 4.1.3). Following the four-month cadence from the June release of 4.1, version 4.2 is projected to arrive in October 2026. Key Initiatives for 2026: According to the 2026 roadmap, upcoming major features likely to be seen in the 4.x line include: * AI & Vector Search: Native vector indexing and AI-assisted analytics. * Lakehouse Enhancements: Full-featured SQL engine for Iceberg (including support for Iceberg V3 and Delete/Update operations) and Paimon. * Real-time Analytics: Automatic tablet splitting for easier scaling and sub-second data freshness. Please note that release dates are estimates and can shift based on development progress. You can monitor the official StarRocks GitHub Releases page for the most up-to-date information. References • release_notes/release-4.0.md • release_notes/release-2.5.md • release_notes/release-3.3.md • release_notes/release-3.4.md • release_notes/release-2.4.md
s
@Rocky 4.1.3 is stable version or 4.1.4?
r
As of late September 2026, version 4.1.5 is the most recent and recommended patch in the 4.1 line, having been released on September 8, 2026. Between 4.1.3 and 4.1.4: * v4.1.4 is a more updated patch than 4.1.3, containing additional bug fixes. * v4.1.5 is currently the latest stable patch for the 4.1 minor version. It includes critical improvements such as reduced lock contention, optimized metadata refresh for Paimon, and various CVE security updates. If you are choosing between 4.1.3 and 4.1.4 for a production environment, you should generally opt for the latest available patch (currently 4.1.5) as StarRocks patch releases are cumulative and focused on stability and bug fixes. Summary of Current Stable Versions (as of Sept 2026): * 4.1.x Line: 4.1.5 (Latest) * 4.0.x Line: 4.0.14 * 3.5.x Line: 3.5.21 (LTS/Stable) References • release_notes/release-4.1.md • release_notes/release-4.0.md • release_notes/release-3.5.md References • release_notes/release-4.0.md • release_notes/release-2.1.md
s
When will be plan of 4.2 @Rocky what are improvement and bug fixes
r
Based on the current StarRocks release cycle and the 2026 Roadmap, version 4.2 is expected to focus on deepening Data Lakehouse capabilities, AI/Vector search, and scaling efficiency. 1. Estimated Release Date StarRocks typically follows a four-month minor version release cycle. * Version 4.1 was released in early June 2026. * Version 4.2 is projected to be released in October 2026. 2. Key Improvements (2026 Roadmap) Major initiatives planned for the 4.x line include: * Lakehouse Full-Featured SQL: * Iceberg Write Enhancements: Support for 
DELETE
, 
UPDATE
, and 
MERGE INTO
directly on Iceberg tables. * Iceberg V3: Full support for the Iceberg V3 specification and Variant data type. * Paimon Support: Native 
INSERT
support, along with Tag/Branch management and Time Travel for Paimon catalogs. * Real-time & Scaling: * Auto Tablet Split: Automatically splitting tablets/buckets to handle larger data volumes under the shared-data architecture. * Multi-Node Tablet Writes: An enhancement specifically tagged for 4.2 (#79405) allows multiple compute nodes to write to the same tablet in shared-data mode, increasing ingestion throughput. * Sub-second Data Freshness: Further optimizations to the real-time storage engine. * AI & Vector Search: * Native Vector Index: Built-in vector indexing to support AI-driven similarity search workloads. * Full-text Inverted Index: Improvements to the built-in inverted index for faster text search. * Performance: * Global Late Materialization: Optimizing query performance by delaying data reading until after filtering. * Fine-grained Tablet Parallelism: Improving concurrency within a single tablet scan. 3. Bug Fixes Specific bug fixes for 4.2 will be finalized upon its release. Currently, the StarRocks team is actively addressing stability issues in the 4.1.x patch line, which will be rolled into 4.2. Common focus areas include: * Memory Safety: Robustness in long-string and JSON processing scenarios. * Catalog Stability: Fixing OAuth2/token-refresh issues for Iceberg REST catalogs. * Query Correctness: Fixing edge cases in 
ARRAY/MAP
constructors and complex aggregate pushdowns. You can follow the live progress of version 4.2 features on the StarRocks 2026 Roadmap (GitHub #67632). References • release_notes/release-4.0.md • release_notes/release-4.1.md • release_notes/release-3.5.md References • release_notes/release-4.0.md • release_notes/release-2.4.md • release_notes/release-3.3.md
s
@Rocky any java version upgrade
r
Yes, there is a significant Java version upgrade planned for version 4.2. Planned Upgrade: JDK 21 (v4.2) For the upcoming v4.2 release (expected October 2026), StarRocks is planning to move the Java baseline from JDK 17 to JDK 21 LTS. * Target Components: This upgrade affects the FE (Frontend), Java-extensions (JNI), build systems, and the runtime environment. * Why the move? * Support Life: JDK 17 is approaching its end of primary support window in late 2026/2027. Moving to JDK 21 provides a support runway through 2029. * Performance: JDK 21 introduces Virtual Threads (Project Loom), which offers future opportunities to optimize FE concurrency and RPC/IO performance. * Ecosystem Compatibility: It ensures better alignment with modern versions of Hadoop (3.4.x), Spark (3.5.x), and Iceberg. Recent History (v3.5 to v4.1) If you are coming from an older version, please note the recent changes: * StarRocks v3.5.x & v4.1.x: These versions currently require JDK 17 or later. Java 8 and Java 11 are no longer supported for these versions. * Configuration: Starting with v3.5, StarRocks no longer uses separate JVM configurations for different JDK versions; all settings are unified in 
JAVA_OPTS
within 
fe.conf
. Recommendation for Upgrading: If you plan to move to 4.2 once it is released, you should prepare your environment by ensuring your nodes are ready to run JDK 21. If you are currently on v3.x or 4.0/4.1, you should already be on JDK 17 at a minimum. References • release_notes/release-3.5.md • deployment/upgrade.md • release_notes/release-3.3.md References • release_notes/release-3.5.md • deployment/upgrade.md • release_notes/release-3.3.md References • release_notes/release-3.5.md • developers/type_checker_xml_configuration.md • deployment/environment_configurations.md • deployment/upgrade.md • sql-reference/sql-functions/JAVA_UDF.md • release_notes/release-3.3.md • release_notes/release-4.0.md
s
@Rocky any doc mentioned 4.2 version is with java21
r
Yes, the plan to upgrade the Java baseline to JDK 21 for version 4.2 is documented in official StarRocks feature proposals and release guides. Documentation & Evidence 1. GitHub Feature Request #74519: Titled "[Feature] Upgrade Java baseline from JDK 17 LTS to JDK 21 LTS (v4.2)", this proposal explicitly targets the v4.2 release for moving the FE, JNI extensions, build system, and runtime environment to Java 21. 2. StarRocks Version Release Guide: The official documentation notes that the "Recommended JDK" for the 4.x line is transitioning to JDK 21. While JDK 17 remains the minimum for 4.1, the 4.2 release is designated as the point where the baseline (minimum requirement) is bumped to 21 to ensure a long support runway as JDK 17 approaches its end-of-life in 2026/2027. What the upgrade includes (per GitHub #74519): * Compile Target: Updating 
pom.xml
and 
build.gradle
to target Java 21. * Runtime Minimum: Raising the 
MIN_JDK_VERSION
check in 
start_fe.sh
to 21. * Ecosystem Alignment: Ensuring compatibility with modern versions of Hadoop (3.4.x), Spark (3.5.x), and Iceberg (1.10.x) which fully support JDK 21. * FE Improvements: Preparing the FE for future performance optimizations using Virtual Threads (JEP 444), which is a core feature of JDK 21. Important Note for 4.1 Users: If you are currently on version 4.1.x, you can continue using JDK 17. However, once you plan to upgrade to v4.2, you will be required to install JDK 21 on your FE and BE nodes to meet the new baseline requirements. References • deployment/deployment_prerequisites.md • release_notes/release-3.5.md • Version Release Guide - StarRocks Docs • GitHub Feature #74519: Upgrade Java baseline to JDK 21 (v4.2) References • deployment/deployment_prerequisites.md • release_notes/release-3.5.md • release_notes/release-3.3.md • release_notes/release-4.0.md • release_notes/release-4.1.md • GitHub Issue #74519: [Feature] Upgrade Java baseline from JDK 17 LTS to JDK 21 LTS (v4.2) • GitHub Issue #67632: StarRocks 2026 Roadmap References • deployment/environment_configurations.md • deployment/deployment_prerequisites.md • release_notes/release-3.5.md • faq/Deploy_faq.md • sql-reference/sql-functions/JAVA_UDF.md • release_notes/release-3.3.md
s
@Rocky if iam using eks do I need to make any changes
r
If you are running StarRocks on Amazon EKS (using the StarRocks Operator or Helm), the transition to version 4.2 and Java 21 is designed to be seamless, but there are a few specific points to check: 1. Docker Images (No manual JDK install) You do not need to manually install JDK 21 on your EKS worker nodes. The official StarRocks Docker images (
starrocks/fe-ubuntu
, 
starrocks/be-ubuntu
, etc.) are self-contained and will come pre-bundled with JDK 21 starting with version 4.2. * Action: Simply update the 
image
tag in your 
StarRocksCluster
CRD or 
values.yaml
file to 
4.2.x
. 2. Update the StarRocks Operator First Before upgrading the cluster to 4.2, ensure your StarRocks Operator is up to date (recommended version v1.10.0 or later). The operator manages the rolling upgrade process and ensures that new Java parameters or environment variables are correctly passed to the new pods. 3. Review 
JAVA_OPTS
in your YAML If you have customized the 
JAVA_OPTS
environment variable in your 
values.yaml
or 
StarRocksCluster
specification, review them: * Remove Old Flags: Ensure there are no legacy garbage collection flags (like CMS) that were deprecated in newer Java versions. * Keep Add-Opens: If you use Java UDFs or External Catalogs (Iceberg, Glue, etc.), you must ensure the following flags remain in your 
be.conf
or `JAVA_OPTS`:
Copy code
bash
    --add-opens=java.base/java.util=ALL-UNNAMED
    --add-opens=java.base/java.nio=ALL-UNNAMED
    --add-opens=java.base/sun.nio.ch=ALL-UNNAMED
Note: The official 4.2 images will likely include these by default, but double-check if you override them. 4. EKS-Specific Considerations * IAM Roles for Service Accounts (IRSA): If you use S3-backed storage (Shared-Data) or Glue Catalogs, your IRSA configurations remain the same. Java 21 is fully compatible with the AWS SDKs used by StarRocks. * Graviton (ARM) Nodes: If you are using Graviton-based instances (e.g., 
m7g
, 
r7g
), StarRocks 4.2 will continue to provide ARM64-compatible images with JDK 21. * Storage (EBS/EFS): No changes are required to your Persistent Volume Claims (PVCs) for the Java upgrade. 5. Recommended Upgrade Step As with any major version or baseline change, it is recommended to test the upgrade in a staging EKS namespace first. Use the Operator's rolling upgrade capability to monitor the health of one FE/BE at a time. References • deployment/sr_operator.md • deployment/helm.md • administration/management/FE_configuration.md • GitHub Issue #74519: [Feature] Upgrade Java baseline from JDK 17 LTS to JDK 21 LTS (v4.2) References • deployment/sr_operator.md • deployment/helm.md • administration/management/FE_configuration.md • deployment/environment_configurations.md • deployment/deployment_prerequisites.md • GitHub Issue #74519: [Feature] Upgrade Java baseline from JDK 17 LTS to JDK 21 LTS (v4.2) • StarRocks on EKS - Best Practices Guide • release_notes/release-3.5.md References • deployment/sr_operator.md • deployment/helm.md • quick_start/helm.md
t
@Rocky OOM is resolved, but many tables have compaction scores that aren't improving, even if I stop all data loading
r
If your compaction scores are not improving even after stopping data loading, it usually means the compaction process is either throttled by configuration, stuck due to resource limits, or the backlog is too large for the current thread allocation. Follow these steps to diagnose and resolve the issue: 1. Check if Compaction is actually running Run the following command to see if there are active compaction tasks:
Copy code
sql
SHOW PROC '/compactions';
* What to look for: If the list is empty or tasks stay in the queue (
in_queue_sec
is high), your BEs are not picking up new tasks. * Shared-data (Cloud Native): Check 
information_schema.be_cloud_native_compactions
for failure reasons. 2. Check Disk Space (The "High Water Mark") StarRocks will stop compaction if a disk reaches a certain usage threshold to prevent full disk failure.
Copy code
sql
SHOW PROC '/be_nodes';
* Check the 
UsedPct
column. If any disk is above 90-95% (controlled by 
storage_flood_stage_usage_percent
in 
be.conf
), compaction may be disabled automatically. You must free up space or add nodes to resume. 3. Increase Compaction Threads If the backlog is large, the default number of threads might be too low to clear it quickly. You can increase these dynamically (no restart required for most): For Duplicate/Aggregate/Unique Key tables:
Copy code
sql
-- Increase threads per disk (Default is usually 1 or 2)
ADMIN SET FRONTEND CONFIG ("cumulative_compaction_num_threads_per_disk" = "4");
ADMIN SET FRONTEND CONFIG ("base_compaction_num_threads_per_disk" = "2");
For Primary Key tables: Primary Key compaction is more memory-intensive.
Copy code
bash
# In be.conf or via curl
update_compaction_num_threads_per_disk = 4
4. Check for "Stuck" Rowsets in BE Logs If a specific tablet has a corrupted rowset or a version gap, compaction might fail repeatedly. Search your BE logs (
<http://be.INFO|be.INFO>
) for error messages:
Copy code
bash
grep -i "compaction" <http://be.INFO|be.INFO> | grep -i "fail"
Common issues include 
version already exists
or 
rs_id not found
. If you see a specific tablet ID failing repeatedly, it may need manual repair. 5. Manually Trigger Compaction (Primary Key Tables) If you are using Primary Key tables, you can try to force a compaction on a specific table to see if it makes progress:
Copy code
sql
-- This syntax is for PK tables to trigger manual compaction
ALTER TABLE <table_name> OPTIMIZE;
6. Verify Memory Limits Since you recently resolved OOM issues, check if your 
compaction_mem_limit
is set too low. If compaction tasks require more memory than allowed, they will be canceled. * Check 
compaction_max_memory_usage_percent
in 
be.conf
(default is usually 80-100% of the process limit, but check if it was manually lowered during your OOM troubleshooting). Summary Checklist: 1. Is Disk Usage < 90%? 2. Are there active tasks in 
SHOW PROC '/compactions'
? 3. Are there "fail to compaction" errors in 
<http://be.INFO|be.INFO>
? 4. Have you increased 
cumulative_compaction_num_threads_per_disk
? References • administration/management/compaction.md • faq/loading/Loading_faq.md • administration/management/BE_configuration.md References • administration/management/compaction.md • faq/loading/Loading_faq.md • knowledge/admin_and_operation/shared_data_compaction.md • administration/management/BE_configuration.md • administration/management/FE_configuration.md
t
@Rocky I see compactions that have run and completed for today, but not for all tables, some compactions only slightly improve compaction score. Disks are less than 10% full currently, each disk can actually fit the entire s3 bucket currently. If I manually run a compaction 
ALTER TABLE table_name COMPACT;
, I don't see the compaction operation even get listed
r
In a Shared-data (Cloud Native) environment on EKS, compaction is orchestrated by the Leader FE and executed by the CN (Compute Nodes). If your manual 
ALTER TABLE ... COMPACT
command isn't appearing in 
SHOW PROC '/compactions'
, or scores aren't dropping, here is how to troubleshoot: 1. Why 
ALTER TABLE ... COMPACT
might not show up If the command returns "OK" but nothing appears in `SHOW PROC '/compactions'`: * Version Compatibility: Manual compaction for Shared-data was introduced in v3.3.0. If you are on an older version, the SQL syntax might be accepted but not implemented for cloud-native tables. * Minimum Score Threshold: By default, StarRocks won't trigger a compaction task if the partition's score is below the threshold set by 
lake_compaction_score_selector_min_score
(default is 10). If your "high" score is actually around 10-15, the FE might decide it's not worth the resource cost. * Check the FE Leader Log: Search the 
fe.log
on the Leader FE for the table name or "compaction" immediately after running the command. It will often log why a task was skipped (e.g., "no partition to compact" or "already has a running transaction"). 2. Why scores "slightly improve" but stay high Compaction in StarRocks is size-tiered. This means it groups files of similar sizes together. * The "Large File" Wall: If you have one massive data file and 20 tiny ones, StarRocks will merge the 20 tiny ones into a medium one. Your score will drop (e.g., from 21 down to 2), but those medium files won't merge with the massive one until you have enough "medium" files to trigger a larger tier merge. * Singleton Delta Limit: The parameter 
max_cumulative_compaction_num_singleton_deltas
(default 100-500) limits how many files can be merged in one go. If you have 2,000 versions, one compaction run will only clear 500, requiring 4 separate runs to fully clear the backlog. 3. Tuning for a Deep Clean Since your disks are empty and you have no load, you can force the system to be more aggressive: On the FE (Global settings):
Copy code
sql
-- Allow more concurrent tasks (default is usually #CNs * 16)
ADMIN SET FRONTEND CONFIG ("lake_compaction_max_tasks" = "100");

-- Lower the threshold to allow cleaning even low-score partitions
ADMIN SET FRONTEND CONFIG ("lake_compaction_score_selector_min_score" = "0");
On the CN/BE (via 
update_configs
or 
be.conf
):
Copy code
sql
-- Increase the number of threads per node
UPDATE information_schema.be_configs SET VALUE = "16" WHERE name = "compact_threads";

-- Speed up large backlogs by merging more files per task
UPDATE information_schema.be_configs SET VALUE = "1000" WHERE name = "max_cumulative_compaction_num_singleton_deltas";
4. Advanced Debugging Commands To see exactly what the FE thinks about your partitions and why it might be ignoring them, run:
Copy code
sql
-- See exactly which partition has the highest MaxCS
SELECT * FROM information_schema.partitions_meta 
WHERE TABLE_NAME = 'your_table_name' 
ORDER BY MAX_CS DESC;

-- Look for hidden errors in recent cloud-native tasks
SELECT * FROM information_schema.be_cloud_native_compactions 
WHERE STATUS != 'OK' 
ORDER BY START_TIME DESC LIMIT 10;
Recommendation: If the score is stuck and manual SQL fails, try a Base Compaction specifically, which forces a merge into the base version:
Copy code
sql
ALTER TABLE <table_name> BASE COMPACT;
References • administration/management/compaction.md • knowledge/admin_and_operation/shared_data_compaction.md • sql-reference/sql-statements/table_bucket_part_index/ALTER_TABLE.md • administration/management/BE_configuration.md
t
@Rocky I've run the base compaction, some times it helps, other times it results in the same exact score Profile: {"sub_task_count":6,"read_local_sec":0,"read_local_mb":0,"read_remote_sec":0,"read_remote_mb":0,"read_segment_count":500,"write_segment_count":0,"write_segment_mb":0,"write_remote_sec":0,"in_queue_sec":2,"score_before":{"avg":166.66666666666666,"p50":0.0,"max":1000.0},"score_after":{"avg":166.66666666666666,"p50":0.0,"max":1000.0},"partial_success":false}
r
The profile you shared indicates a "no-op" compaction, where the system did the work of reading data but failed to produce any new merged files (write = 0) or reduce the version count. Looking at your profile: * `read_segment_count: 500`: This is the exact default limit for 
max_cumulative_compaction_num_singleton_deltas
. Your compaction is hitting the maximum allowed segments per task. * `write_segment_count: 0`: No new data files were written to S3. * 
score_after
== `score_before`: The "MaxCS" (Max Compaction Score) did not budge. This pattern typically happens for one of three reasons in StarRocks Shared-data clusters: 1. The "Size-Tiered" Trap StarRocks groups files into "tiers" based on size (e.g., <1MB, 1-10MB, 10-100MB). * The Issue: Compaction only merges files within the same tier. If you have 1,000 tiny files, it will merge 500 of them into one "medium" file. * Why the score doesn't change: The Compaction Score is often dominated by the tier with the most files. If you merge 500 tiny files into 1 medium file, you still have 500 tiny files left in that tier. The 
MaxCS
remains high because the "backlog" in that specific size group is still there. * Why 
write_segment_count
is 0: If you are using Primary Key tables and the compaction task only processed delete bitmaps or found that all rows in those 500 segments were already deleted/overwritten, it may result in 0 new data segments. 2. Primary Key Table: Persistent Index Issues If this is a Primary Key table, compaction requires the "Persistent Index" to be healthy. * If the index is corrupted or the CN node running the compaction is under extreme memory pressure (which you mentioned having recently), the compaction task might "finish" without actually committing the merge to storage. * Check: Run 
SHOW INDEX FROM table_name;
and check the BE logs for 
fail to load index
or 
index checksum mismatch
. 3. The "Max Version" Limit If a single tablet has an extremely high version count (near 1,000), and the compaction task is limited to merging 500 (
max_cumulative_compaction_num_singleton_deltas = 500
), it only solves half the problem. If the task fails to commit for any reason (e.g., S3 timeout during the final commit phase), the score won't refresh. ────────── Recommended Actions to "Force" Progress: 1. Increase the "Bulk" merge limit (CN side) Since you have a score of 1000, you need to merge more than 500 files at once to see the score drop significantly.
Copy code
sql
-- Allow a single task to merge up to 1000 files instead of 500
UPDATE information_schema.be_configs SET VALUE = "1000" WHERE name = "max_cumulative_compaction_num_singleton_deltas";
UPDATE information_schema.be_configs SET VALUE = "1000" WHERE name = "lake_pk_compaction_max_input_rowsets";
2. Check for "Stuck" Transactions Compaction is a transaction. If there is a "stale" or "failed" transaction holding a lock on those rowsets, new compactions will do the work but fail to "Publish" the result.
Copy code
sql
-- Check for any "PREPARED" or "COMMITTED" transactions that are very old
SHOW PROC '/transactions';
3. Inspect the BE/CN Logs for "Write" Errors Since 
write_segment_count
is 0, search your CN/BE logs for the TxnID from your 
SHOW PROC '/compactions'
output:
Copy code
bash
grep "TxnID_from_profile" <http://be.INFO|be.INFO>
Look specifically for: * 
fail to upload to s3
* 
tablet is under compaction
* 
too many versions
4. Temporary "Aggressive" FE Selection Force the FE to pick this table more often by lowering the selection threshold:
Copy code
sql
ADMIN SET FRONTEND CONFIG ("lake_compaction_score_selector_min_score" = "0");
Question for you: Is this a Primary Key table, and did you recently perform a massive DELETE or UPDATE operation? This often causes the "0 write" behavior because the compaction is effectively just cleaning up deleted records. References
t
sorry for the noise, I had thought I opened up my previous thread
w
@Rocky is there a 4.2 list of features?
r
As of October 2026, StarRocks v4.2 is in the late stages of development (Release Candidate phase). While the official GA release notes are typically published upon the full stable release, the following features have been tagged for v4.2 in the community roadmap and GitHub repository: 1. Data Lakehouse & Iceberg Enhancements * Support for 
MERGE INTO
on Iceberg V2: This is a major update allowing Merge-On-Read (MOR) operations directly on Iceberg V2 tables. This enables more efficient updates and deletes for lakehouse workloads without requiring a full table rewrite. * Iceberg Row-Delta Commit Path: Integration of a native commit path for row-level changes in Iceberg. 2. AI & Vector Search * Vector Search Framework: Significant updates to the internal vector search framework are slated for 4.2, improving the interface for AI-driven applications and similarity searches. 3. Observability & Management * JSON Audit Logs: A new FE configuration 
audit_log_json_format
allows StarRocks to output audit logs in structured JSON, making it easier to integrate with modern logging stacks like ELK or Datadog. * Native Linux Packaging: Official support for Debian (
.deb
) and RPM packages is being introduced to simplify standard Linux installations (replacing the manual tarball extraction method for some users). * Online `OPTIMIZE TABLE`: Enhancements to the 
ALTER TABLE ... OPTIMIZE
command to reduce write-blocking during background maintenance tasks. 4. System & Performance * StarOS 4.2: The underlying cloud-native storage management layer (StarOS) is being upgraded to version 4.2, which includes optimizations for shared-data cluster stability and S3 object management. * CTE Optimization: New session variables to control the "force reuse" of Common Table Expressions (CTEs) to speed up the optimizer for extremely complex queries. How to verify Since 4.2 is nearing release, you can see the latest progress by: 1. Checking the StarRocks/starrocks GitHub Issues with the label 
version:4.2
. 2. Looking at the 
branch-4.2
or 
main
branch code in the official repository. 3. Monitoring the official StarRocks Blog for the upcoming "StarRocks 4.2: What's New" announcement.
Open in Slack
PreviousNext
Powered by