CUBRID 11.4 has been released.
CUBRID 11.4 is the latest stable version that includes new features, significant changes, and improvements.
CUBRID 11.4 introduces:
1) PL/CSQL support for Oracle compatibility
2) Addition of HASH JOIN for large-scale processing
3) Performance improvements through optimizer and index processing enhancements
4) Performance boost in data recovery through parallel processing
5) Performance enhancement by expanding result caching
6) Improved data dump performance
7) Addition of memory monitoring features
8) Enhanced access control functionality
9) Improved user convenience for backup and recovery operations
These updates bring not only new features but also various performance improvements across multiple areas.
1. Changes
* New features
- HASH JOIN Support
- Support for Changing Owner of SERIAL SQL Syntax
- Support for SQL Statements to Add/Remove Members from a Created Group
- Extended Query Cache Support
- Added LEADING Hint to Specify Join Order
- Added Option to Delete Specific Plans in cubrid plandump
- Added Support for Separate loadjava Option for JNI-Based JavaSP
- Added Memory Monitoring Featured utility
- Added support for the diagdb -d 9 option to dump data from a specific class (table)
- Added Parameter to Set the Size of Data Sent to Clients in the Broker
- Added Parameter to Set ACL(Access Control) for Each Broker
- Support for Automatic Restart of Non-HA DB Server After Abnormal Termination
- Support for Script to Restore Backup as a New Database
* Specification Changes
- Limit Maximum Number of Characters for CHAR Type to 2,048 (Previous Limit: 268,435,456)
- LOB Locator Path Changed to Relative Path
- Restriction on Use of Analytic Functions in UPDATE JOIN Queries
- Restrict Use of for update clause on System Tables and Views
- Data Type Consistency Check Disabled During View Creation
- Allow Use of NULL in SELECT Clause of View Query Specification
- Disallow Use of Both AUTO_INCREMENT and DEFAULT in TABLE Modification and Creation.
- Raise Error When Adding New Columns in ALTER INDEX … REBUILD
- Raise Error When ROWNUM Value Exceeds NUMERIC Type Range
- Recognize Session Commands While Writing SQL or PL/CSQL Statements(like create, body and etc) in csql Interactive Mode
- Suppress Plan Information Output Before Deletion in cubrid plandump -d
- Raise Error When ha_mode is on and myhost is Included in ha_node_list
- Column Name Changed in System Table db_serial (att_name → attr_name)
* Improvements (Including Performance Enhancements)
- Performance Improvement with Caching for Correlated Subqueries
- Enhancement of Sort-Limit Optimization When Bind Variables or Expressions Are Used in the LIMIT Clause
- Improved Automatic Restart of JavaSP Server When JNI Program Encounters Segmentation Fault
- Enhanced String Functionality to Support Multiple Character Encodings (euckr, utf8)
- Removed Length Limitation on WHERE Clause When Creating Filter Indexes
- Improved Index Scan Performance by Optimizing midxkey.buf Size Calculation for Multi-Column Indexes
- Optimized Index Scan Performance by Reducing Unnecessary Calculations and Comparisons
- Performance Improvement Through Overall Optimizer Enhancements
- Improvement in Stored Procedure Execution Plans
- Improvement in Concurrency by Preventing Unnecessary X-LOCKs on Unneeded Rows
- Improvement in Index Scanning for LIKE Queries Using JavaSP Functions
- Improvement in Range Index Scanning for Queries with <= and >= Conditions on Function Indexes
- Improvement to unloaddb in Databases Without Tables and Views to Extract Other Schemas(user, serial, sp, server, synonym, grant)
- Performance Improvement in unloaddb
- Improvement in Clarity of csql Environment Variable Names
- Improved the output of the cubrid spacedb command
- Improvement in cubrid spacedb Output
- Error Occurrence for Incorrect ACL Settings for Improved User Awareness
- Improvement to Allow TLS v1.2 Clients to Connect to CMS(CUBRID Manager Server)
- Improvement to Prevent Zombie Processes When Broker/CAS Processes are Terminated Through CMS
- Improvement to CMS getlogfileinfo() API to Return SQL Log File Information Only Once
- Improvement to cms ha_status() API to Display Replica Node Status in HA Environments with Master, Slave, and Replica Configuration
- Clear Error Message Display for Incorrect Settings in ha_node_list and ha_replica_list for Improved User Awareness
- Clear Error Message Display for Failover and Failback Occurrences to Improve User Awareness
- Improvement to Prevent Incorrect Data Mismatch Error Message from Appearing When restoreslave is Executed on Slave Node
- Improvements to User Hosts (cubrid_host.conf) for User Convenience
- Improvements to DDL Audit Logs
- Performance Improvement in REDO Recovery with Parallel Threads
- Simplified Lock Information Logged in dbname_latest.event File During Deadlock
- Performance Improvement in time_format() and date_format()
- Query Performance Improvement with TRACE Enabled
- Performance Improvement by Removing Unnecessary Length Checks on String Types
- Memory Usage Improvement by Limiting Lock Resource Retention to lock_escalation Setting
- Improved User Awareness with Error Output for Invalid optimization_level Setting
- Improvement in Checking DB Volume and Log Creation Times
- Improvement in User Awareness by Adding Fence Key Information to SHOW INDEX CAPACITY and diagdb
* Bug Fixes
- Fixed Hidden Column Handling in Scalar Subqueries with ORDER BY
- Fixed an issue where referencing a remote server object without specifying the user schema could fail to resolve to the correct object if an object with the same name exists in another user.
- Fixed an issue where View Merging of Inline Views and Scalar Subqueries Containing ORDER BY Clauses, Where ROWNUM Was Incorrectly Rewritten as ORDER_BY_NUM()
- Corrected an Issue When oracle_style_empty_string=yes is Set and The Result of The Expression ‘null || string’ Was Incorrectly Returned as NULL
- Fixed NULL Output in REPLACE Function When oracle_style_empty_string=yes
- Fixed Incorrect Results in Covered Index Scans Using Function Indexes
- Fixed Segment Fault Errors in CTE Queries with Always False (or NULL) Conditions
- Fixed Segment Fault Errors in UNION ALL Queries Using Multiple Tables and Serial Next Values
- Fixed Segment Fault Errors in Queries Using Constants Over 255 Characters Without Aliases
- Fixed Errors During View Merging with OUTER JOIN
- Fixed Displaying Both User and Public Views With the Same Name in SHOW CREATE VIEW
- Fixed Missing Error Message and NULL Return in repeat() Function Exceeding string_max_size_bytes Setting
- Fixed Incorrect Table Positioning During INNER JOIN Removal in Query Optimization
- Fixed Errors with Host Variables in PREPARE Statements Using Certain Built-in Functions
- Fixed Missing User Information Deletion in System Catalog Tables When Dropping Users by The DROP USER Clause
- Fixed Oracle-Style LEFT OUTER JOIN Not Rewriting to INNER JOIN When Using Host Variables in SP Call or WHERE Clause
- Fixed Incorrect Results in Subquery with ORDERBY_NUM() During View Merging
- Fixed for Tables Without Join Relations Being Recognized as Join Tables During Query Optimization
- Fixed ROWNUM Values Being Printed as 0 When Using Equal Condition (=) in WHERE Clause and the Same Column in ORDER BY
- Fixed Errors That Occur When INNER JOIN and Oracle-Style OUTER JOIN Are Used Together and JOIN Condition Placement in WHERE Clause
- Corrected Incorrect Results in INSERT INTO tbl … SELECT … FROM View … ON DUPLICATE KEY UPDATEI Query
- Corrected Incorrect Column Order Numbers in Analytical Functions Inside Views
- Resolved a Case Where Range Conditions Were Not Recognized as Range Items When Using The Pipe Operator (||)
- Fixed An Error That Occurred When Handling NULL in Host Variables of DBLINK Statements
- Fixed Issue Where Korean Characters Were Corrupted When Inserting via Host Variables Using DBLINK on EUC-KR Oracle DB
- Resolved an Issue With Using DATETIMELTZ as a DATETIME Parameter in JavaSP
- Fixed An Error That Occured Due to Space Between Declared Java Method Name and Opening Parenthesis in JavaSP
- Fixed an Error that Occurred When Using Columns of Tables with Unsupported Types as Arguments in JavaSP
- Fixed The Cannot allocate query entry any more Error in JavaSP
- Fixed The Exit Code Not Being Set to 3 When an Error Occurs During loaddb Execution
- Fixed Incorrect count(*) Value Returned After Executing loaddb with --no-logging
- Fixed Error That Occurred When Processing Serials Without Schema Names During loaddb Execution
- Fixed an Issue Where loaddb Would Continue Execution Even After an Error Occurred
- Resolved an Error That Occurred When Processing a Synonym for System Tables During loaddb Execution
- Corrected an Issue Where the Schema Name Was Incorrectly Stored or Missing for Trigger System-Table Columns (condition, action_definition) During loaddb Execution Using unloaddb Files From Version 11.4 or Earlier
- Addressed a Case Where Reverse Unique-Index Comments Were Missing During unloaddb Execution
- Corrected an Issue Where the auto_increment Value Was Reset to 1 for PK Columns With auto_increment Set During unloaddb Execution
- Resolved an Issue Where the dbname_schema_uk File Was Created Even When No Unique Index Existed During unloaddb --split-schema-files Execution
- Addressed a Case Where the ALTER SERIAL Statement Was Missing When Executing unloaddb With the -i (--input-class-only) Option
- Resolved an Error That Occurred When current_val and max_val of a Serial Were Identical During unloaddb Execution
- Corrected an Issue Where an Infinite Loop Occurred When Processing JSON Data Over 1MB During unloaddb Execution
- Addressed a Case Where, When Running unloaddb as a Regular User, Schema Names Identical to the User Were Not Removed From the query_spec of Views
- Resolved a Case Where Schema Names for Serials and Triggers Were Missing When Executing unloaddb as DBA User
- Corrected an Issue That Occured When Running unloaddb as a DBA User Without the --as-dba Option, Where Procedure Privileges Granted by Other Users (e.g., GRANT ... ON PROCEDURE) Were Incorrectly Included in the Output
- Corrected an Issue That Occured When Running unloaddb as a Regular User, Where Schema Names Identical to the User Were Not Removed From the Condition and action_definition of Triggers
- Fixed for Unnecessary Archive Log Volumes Being Created During backupdb Execution
- Fixed an Issue Where flashback Did Not Terminate Properly on Forced Exit (Ctrl+C)
- Resolved a Case Where backupdb Could Hang in Certain Situations
- Fixed a Potential Memory Leak In the Processing of addBatch() and executeBatch() in Broker/CAS/CMS
- Corrected an Incorrect Output From the getTransactionInfo() Function Call
- Addressed a Case Where Replication Logs Were Not Deleted on a Replica Node When ha_replica_delay Was Set to 60 Seconds or More in an HA Environment With Master, Slave, and Replica Nodes
- Resolved an Issue Where the SQL Log File Was Not Automatically Deleted Under Certain Conditions During applylogdb Execution
For more details, please refer to the CUBRID 11.4 change logs(https://github.com/CUBRID/cubrid/releases/tag/v11.4) and online manuals (https://www.cubrid.org/manual/en/11.4/release_note/index.html)
2. Precautions
- When using heterogeneous(Oracle/MySQL) DBLINK, you must use ODBC Drvier for Unicode of Oracle/MySQL.
3. Environmental Requirements
- Linux: 64bit, kernel 2.4 or higher, Windows: Windows7 or higher (Windows 7 requires the latest security updates)
- JRE 1.8 or higher is required when using JSP or CUBRID Admin/CUBRID Manager
4. Download
- Product download : https://www.cubrid.org/downloads
- Product source code : https://github.com/CUBRID/cubrid/tree/v11.4
