Features & Improvements

• Others: Remove risky RPC interfaces.
• Others: Remove JEXL functions.
• Others: Check naming when creating a Pipe.
• Others: Change the default client RPC service address to 127.0.0.1.
• Others: Modify code to make internal services bind to the value specified by dn_internal_address instead of the default address.

Bugs

• Fixed an issue in the partition table automatic deletion logic: Only TTL configured at the DB level takes effect, using the maximum TTL value within the DB.

Full Changelog: v1.3.6...v1.3.7

Features & Improvements

• Others: Remove risky RPC interfaces.
• Others: Remove JEXL functions.
• Others: Check naming when creating a Pipe.
• Others: Change the default client RPC service address to 127.0.0.1.
• Others: Modify code to make internal services bind to the value specified by dn_internal_address instead of the default address.

Bugs

• Fixed an issue in the partition table automatic deletion logic: Only TTL configured at the DB level takes effect, using the maximum TTL value within the DB.

Full Changelog: v2.0.6...v2.0.7

Features & Improvements

• Data Query: Supports table model query and write-back functions.
• Data Query: The MATCH RECOGNIZE feature for table models supports aggregate functions to capture continuous data for analysis and calculation.
• Data Query: Added built-in scalar functions bit manipulation functions for table models.
• Data Query: Added pushdown-enabled EXTRACT time function for table models.
• Others: Fixed security vulnerabilities CVE-2025-12183, CVE-2025-66566 and CVE-2025-11226.
• ...

Bugs

• Fixed the issue where executing queries with CASE WHEN throws an internal server error.
• Fixed the issue where creating a view succeeds but querying the view fails when the tree model database name is the same as the device name in tree-to-table scenarios.
• Fixed the issue where the automatic database creation function fails subsequently if a database is deleted after successful table model data synchronization, and the database is not restarted.
• Fixed the issue where modifying the password of a user that failed to be created throws an internal server error.
• Fixed the issue where an error is thrown when using the IF PERMITTED clause to modify sequence encoding and compression settings.
• Fixed the issue where querying BLOB data types via the tree model RESTAPI interface /rest/v2/query returns garbled results.
• Fixed the issue where incorrect exception information is displayed when checking pipe status after creating a pipe in a table model that only contains tree-to-table views.
• Fixed the issue where the number of server connections keeps increasing when using a user without MAINTAIN permission (non-root user in table models) in a Session.
• ...

Full Changelog: v2.0.5...v2.0.6

Features & Improvements

• Data Query: Optimized query performance for multiple scenarios, including the multi-series Last query.
• Data Query: Added the FastLastQuery interface to the Java SDK, supporting more efficient Last query operations.
• Data Query: Adjusted the tree-model fetchSchema to return data in segmented streaming mode, improving response speed in large-data-volume scenarios.
• Storage Management: Optimized memory management to avoid memory leak risks and ensure long-term stable system operation.
• Storage Management: Optimized the file compaction mechanism to enhance compaction efficiency and reduce system storage resource consumption.
• Data Synchronization: Optimized Pipe SQL parameter configurations to support specifying asynchronous loading methods.
• Data Synchronization: Added a syntactic sugar feature that automatically splits full-volume Pipe creation SQL into real-time synchronization and historical synchronization.
• System Management: Added a global configuration item for data type compression methods, supporting on-demand adjustment of storage compression strategies.
• Others: Fixed security vulnerabilities CVE-2025-12183, CVE-2025-66566 and CVE-2025-11226.
• ...

Bugs

• Fixed the issue where selecting discontinuous sequence files for the settle task from the sequence zone under device-level TTL would trigger an overlap error during execution.
• Fixed the issue where running the pattern_match custom function in tree-model SQL (3C3D deployment) would throw an error indicating "tsDataType" is null.
• Fixed the issue where TTL would continuously write data into the modification cache in fast compaction mode, resulting in slow compaction speed.
• Fixed the stack overflow issue that occurred when performing value filter queries with a large number of devices without adding the align by device clause.
• Fixed the issue where loading TsFiles with empty value chunks across time partitions would generate TsFiles with misaligned chunks.
• Fixed the incorrect processing logic of TransformOperator for empty TsBlocks.
• Fixed the issue where the number of server connections would keep increasing when using a user without MAINTAIN permission in a Session.
• Fixed the issue where Load would ignore mods import under certain special circumstances (e.g., mismatched aligned types or inconsistent table schemas).
• Fixed the issue where data that had been deleted would be resynchronized to the receiver if mod files existed when Pipe pattern was configured.
• Fixed the issue where info logs related to TTL checks would still be printed during compaction scheduling even when TTL was not configured.
• ...

Full Changelog: v1.3.5...v1.3.6

Improvements

• Storage Module: Optimize user password encryption by changing the algorithm to SHA-256
• ...

Bugs

• Fixed the array out-of-bounds exception in cross-Region aggregate query with align by device
• Fixed the error in aggregate queries with order by time + align by device for single or multiple devices spanning across Regions
• Fixed the "out of memory" error in query with large TEXT objects + align by device + order by time
• Fixed the issue where the last query with where time > X and time < X was not applied to partitioned table fetching, resulting in failure to hit PartitionCache
• Fixed the error in using change_point function + align by device when single device data spans multiple Regions
• Fixed the OOM (Out of Memory) issue caused by the asynchronous submission queue of pipeTransferFile
• Fixed the memory leak issue in pipe password detection
• For the pipe receiver and during Load, change the behavior of'clear schema cache' to 'update last cache'
• Optimize the loading speed of TsFile when there are a large number of measuring points and the mods file contains deletions of a large number of measuring points
• ...

Full Changelog: v1.3.4-1...v1.3.5

Features & Improvements

• Data Query: Support manual creation of tree-to-table views
• Data Query: Table model adds window functions
• Data Query: Table model adds the aggregate function approx_most_frequent
• Data Query: Table model JOIN function expansion, supporting LEFT & RIGHT JOIN, ASOF LEFT JOIN
• Data Query: Table model supports row pattern recognition, enabling capture of continuous data for analysis and calculation
• Storage Management: Table model adds multiple system tables, such as: VIEWS (table view information), MODELS (model information), etc.
• AI Management: AINode adds two built-in models: Timer-XL and Timer-Sundial
• AI Management: AINode supports inference functions for tree models and table models
• ...

Bugs

• Fixed the issue where the where time > X and time < X in last query was not applied to partitioned table fetching, failing to hit PartitionCache
• Fixed the error in aggregate query with order by time + align by device for single device across regions
• Fixed the error in aggregate query with order by time + align by device for multiple devices across regions
• Data synchronization batching supports sink.batch.max-delay-ms, providing more fine-grained cache time control
• Fixed the potential NPE issue when the tablet contains null values during data synchronization with sink format=tsfile
• Fixed the incorrect processing result of pipe processor when writing tablets with null values with duplicate timestamps
• Fixed the issue of tree model load updating table model schema cache
• Fixed the issue where disruptor queue close could not be completed, which might cause the DN stop script to get stuck
• Fixed multiple issues of data synchronization in normal scenarios, restart scenarios, and deletion scenarios
• Fixed the time formatting issue when exporting table model to csv
• ...

New Contributors

• @jtmer made their first contribution in #15468
• @Young-Leo made their first contribution in #14423

Full Changelog: v2.0.4...v2.0.5

Features & Improvements

• Data Query: Added user-defined table functions (UDTF) and various built-in table functions to the table model.
• Data Query: Added support for ASOF INNER JOIN on time columns in the table model.
• Data Query: Added the approximate aggregation function approx_count_distinct to the table model.
• Stream Processing: Added support for asynchronous loading of TsFile through SQL.
• System Management: Added support for disaster recovery load balancing strategy in replica selection during scaling down.
• System Management: Adapted to Windows Server 2025.
• Scripts and Tools: Categorized and organized script tools, and separated Windows-specific scripts.
• ...

Bugs

• Fixed the memory leak issue in the WAL compression buffer.
• Fixed the issue where the async connector gets stuck after running for a long time.
• Fixed the issue where data subscription cannot be terminated after using data export scripts.
• Fixed the issue where pipe restarts frequently due to insertnode and resource management memory problems under memory pressure.
• Fixed the NPE issue triggered during memory statistics estimation on the data synchronization receiver end.
• Fixed the error when configuring ConsumerConstant.NODE_URLS_KEY as a cluster address while using SubscriptionPullConsumer to consume data.
• Fixed the deadlock issue on DN startup caused by concurrent agent pipe metadata fetching and CN metadata pushing.
• ...

New Contributors

• @q977734161 made their first contribution in #15322

Full Changelog: v2.0.3...v2.0.4

Features & Improvements

• Data Query: Added new aggregate function count_if and scalar functions greatest / least to the table model.
• Data Query: Significantly improved the performance of full-table count(*) queries in the table model.
• AI Management: Added timestamps to the results returned by AINode.
• System Management: Optimized the performance of the table model's metadata module.
• System Management: Enabled the table model to actively listens and loads TsFile.
• System Management: Added support for TsBlock deserialization in the Python and Go client query interfaces.
• Ecosystem Integration: Expanded the table model's ecosystem to integrate with Spark.
• Scripts and Tools: The import-schema and export-schema scripts now support importing and exporting metadata for the table model.
• ...

Bugs

• Fixed the issue where a single write request exceeding the total size of the WAL queue caused write queries to hang.
• Fixed the issue where the receiver experienced OOM (Out of Memory) after resuming synchronization following a long period of inactivity.
• Fixed the issue where repeatedly setting TTL for DB and Table led to inserted data being unqueryable and returning an empty list.
• Fixed the issue where a regular user with create+insert permissions on a table encountered exceptions when loading tsfile.
• Fixed the issue where passwords were logged when SessionPool getSession timed out.
• Fixed the issue in the Go client tree model query interface where the absence of a check for the Time column led to an "index out of range [-1]" error when retrieving Time column data.
• Fixed the issue where distinct hits aggregate pushdown optimization and is used with group by date_bin, causing execution exceptions in aggregate queries.
• Fixed the issue of whitespace characters at the beginning and end of port and directory address parameters in the configuration file.
• Fixed the issue where setting the maximum number of concurrent RPC clients less than the number of CPU threads caused DN startup failure.
• Fixed the issue where using a template, after activation, writing to extended columns, and then creating a pipe, caused the series under the device to double.
• Fixed the issue where metadata synchronization, creating a pipe after a template, caused the series to double when using show timeseries.
• Fixed the issue where a regular user with INSERT permissions encountered exceptions when exporting metadata using export-schema.sh.
• ...

Full Changelog: v2.0.2-1...v2.0.3

This is a bug-fix version of 2.0.2

• Fix the bug that will remove the data partition table by mistake in case of us/ns time precision and using ttl

This is a bug-fix version of 1.3.4

• Fix the bug that will remove the data partition table by mistake in case of us/ns time precision and using ttl