Skip to content

V2.4.0-release

The main changes to the V2.4.0-release version are as follows:

  1. Distributed computing performance improvement:

    • hasAggs/having/sort/limit/Stable function/ optimization is pushed down to DN, and the performance is improved nearly a hundredfold.
    • ProcLock segmented lock optimization for distributed processes to reduce lock conflicts and improve execution efficiency by about 5 times.
    • execute plan optimization: if the number of redistributed nodes is 1, skip the remote operator and simplify the execution plan.

  2. 2PC transaction optimization:

    • performance optimization of deadlock checking: batch acquisition of gxid; traversal lookup gxid is modified to binary lookup, and the number of try rounds is increased.
    • 2PC adds protection mode to prevent partial submission.
    • drop database adds a prepare process to ensure that there is no residual node failure.

  3. Enhanced ease of use:

    • added pooler multithreaded log feature to facilitate problem analysis.
    • GTM logs are optimized. Logs are aggregated into one file, and active logs are automatically split to facilitate users to view them at any time.
    • support gtm_ctl-l to specify the log file path to start, making it easy for users to manage logs.

  4. The new features support:

    • self-developed partition table pruning, adding IN syntax pruning to speed up pruning in advance.
    • data shuffle supports user-defined functions.
    • allows you to delete columns of a primary table with distributed foreign key constraints.