LightDB 22.2正式發佈

本版本主要特性包括：

1、引入實驗性特性oracle pl/sql存儲過程兼容，兼容絕大部分oracle pl/sql一份代碼可以同時運行於oracle和lightdb

2、lightdb jdbc支持autocommit=true時，fetchsize生效

3、lightdb優化器提示進一步完善，且語義兼容oracle

4、原生分佈式DDL語法，和greenplum一樣

5、支持lightdb實時同步oracle，ltdts_recvlogical，詳細使用後續文章專門更新。本版本TPS可到10000筆每秒，下個版本預計可到4-5萬筆每秒。

6、lt_dump支持cascade，可極大的提高性能。

7、sysdate/systimestamp支持語句級一致性，oracle語義一致

8、merge into支持分佈式版

9、支持龍芯cpu、rocky linux。

在最新版的性能測試中，單機TB級容量下，2.6k*2億文檔，lightdb全文檢索性能高於es 6.x版本，且實時索引。在金融業務系統ETF TA、自建TA中，相同數據量、業務量下，某SZ頭部公募基金性能選型測試中，lightdb性能遠高於某base和某號稱全自研的國產第一數據庫。

目前，lightdb正在BJ絕對頭部券商現場業務測試，SH、合肥中部券商業務測試。

詳細特性如下：

E.1. Release 13.3-22.2

Release date: 2022-06-30

• Support patch for quick upgrade, see lt_patch.

• Add PL/oraSQL loadable procedural language, see plorasql

• lt_hint_plan Enhancements

  • Add three new hints: use_hash,use_nl,full, semantics are consistent with Oracle.

  • The table name in hint is case-insensitive.

  • Hint parallel default is hard and supports no table specification.

  • Hints can be applied to the tables of the subquery when the subquery has only one table.

    EXPLAIN (COSTS false) select /*+use_hash(t1 y)*/* from t1,(select * from t2) y where t1.id=y.id;
    
             QUERY PLAN             
    ------------------------------------
    Hash Join
    Hash Cond: (t1.id = t2.id)
    ->  Seq Scan on t1 @"lt#0"
    ->  Hash
            ->  Seq Scan on t2 @"lt#1"
    (5 rows)
    

• Support User-defined type priority for better function and operator match, see type compatible.

• Support CREATE TABLE DISTRIBUTED BY to create distribute table(depends on extension canopy), see CREATE TABLE.

  create table dist1(a int, b int) distributed by hash(a) shard_count(4);
  

• Auto-explain execution plan logs are printed in a separate log file (*.slow) using JSON format.

• Add extension wal2sql to support Real-time synchronization to Oracle.

• lt_dump support --recreate-schema to generate drop schema(use CASCADE mode) command for better performance, see lt_dump.

• Client Applications and libpq Support LightDB environment variables.

• sysdateand systimestamp indicates the statement start time.

  lightdb@postgres=# begin;
  BEGIN
  lightdb@postgres=*# select sysdate from dual;
      sysdate       
  ---------------------
  2022-06-28 19:44:15
  (1 row)
  
  lightdb@postgres=*# select sysdate from dual;
      sysdate       
  ---------------------
  2022-06-28 19:44:16
  (1 row)
  
  lightdb@postgres=*# end;
  COMMIT
  

  For details, see datetime.

• When install with high availability, enable replication slot default.

• Support Order-Preserving Encryption ,see lt_ope

• The following platform support is added:

  Architecture	OS
  LoongArch	Kylin V10(SP1)
  x86_64	rockylinux
  aarch64	rockylinux

• Add cluster control tool lightdb_service.py to facilitate the overall start and stop of services in high availability and distributed environments.

  start cluster:

  lightdb_service.py -c start
  

  stop cluster:

  lightdb_service.py -c stop
  

• The value 0 in PWR is display as empty, for details, see lt_profile.

• In non-distributed installation, the canopy extension not installed.

• When keepalived and ltcluster are restarted, new log files are generated to avoid confusion with old logs.

• lt_initdb does not restrict the order of options.

• Fix cannot assign XIDs during a parallel operation error occurs occasionally during parallel query in lt_show_plans.

• Release the JDBC driver library based on PostgreSQL.

  • maven repository

    <dependency>
        <groupId>io.github.hslightdb</groupId>
        <artifactId>lightdb-jdbc</artifactId>
        <version>42.2.30</version>
    </dependency>
    

  We fixed some of the issues and enhanced data type compatibility:

  • fix：fetchSize does not take effect in automatic commit mode, causing the whole data to be loaded into the memory and causing out-of-memory problem.

  • support oracle anonymous blocks syntax (see plorasql ).

  • support to_number() function compatible with oracle.

  • support numeric data type convert to Java Long.

  • support char(1) data type convert to Java Character.

  • fix loss of precision when Java setDouble operates on numeric field.

• lt_cron supports linux command.

  SELECT cron.schedule('dayly-touch', '59 23 * * *', 'rm -rf $LTDATA/log/*', 'next', '8', 'linux');
  

• Built-in tool application lvs for load balancing in distributed environment wiith multi-CN.

• Built-in tool application postgresrest.

• Changed file naming rules in the installation directory and instance directory.

• The extension objects in public schema are all moved to the lt_catalog schema.

• RPM dependencies are built into the installation package.

• Support merge into in distributed environment.

詳見lightdb官方網站www.hs.net/lightdb，lightdb官方文檔網站www.light-pg.com。