MogDB 3.0.2

1. Version Description

MogDB 3.0.2 is a patch version of MogDB 3.0.0, which is released on September 16 of 2022. It mainly involves bug fixes based on MogDB 3.0.1.

2. Function Modifications

• Version upgrade optimization: support direct upgrade from v2.0 to v3.0 (without the need to upgrade MogDB from v2.0 to v2.1)

• The gsql '\l' command adds the function of displaying the database.

3. Modified Defects

• Fix the defect that the database frequently prints warning log after the undo_retention_time parameter is enabled.

• Fix the defect that an error is reported once gs_global_chain cannot be found during vacuum or analyze execution after version upgrade.

• Fix the defect that the data in the new tablespace of the primary server cannot be synchronized to the standby server in the primary/standby stream replication.

• Fix the defect that an error is reported when log_min_messages is set to debug5 during the operation of CREATE or SELECT.

• Fix the defect that "error parameter when release fake relation" is continuously reported when the fields of partition table is updated.

• Optimize the insertion of the xlog record in the x86 system.

• Fix the defect that the ustore zone overflows even if the ustore table is not used when the pressure of the short link service is high.

• Fix the defect that the tuples following the cluster command are disordered due to row-level compression.

• Modify the function of not using asynchronous commit for transactions in vacuum.

• Fix the hang problem probably caused by enabling thread pool.

• Fix the hang problem probably caused because no trace data is written during the gstrace stop operation.

• Fix the defect of repeatedly parsing data after restarting an application when an LSN error is reported in multi-transaction scenarios after parallel decoding is enabled.

• Fix the coredump probably caused by asynchronous submission.

• Fix the database coredump caused by manually adding row lock to the view.

• Fix the coredump caused by parallel replay or extreme recovery in page compression scenarios.

• Fix the coredump caused by frequent creation and deletion of a database in the primary database and the database is connected by the standby database using gsql in the primary/standby deployment scenarios.

• Fix the coredump caused by creating a domain with the same name after dropping table of type and using the field in a package.

4. Synchronized with openGauss 3.0.1 Fixes

• Resolve the issue that threshold has been cleaned but is not automatically cleaned during partition table autovacuum.
• Fix logical decode core.
• Fix deserialize func and view.
• Fix gs_dump.
• Fix ustore bug in publication.
• DFX: add exit log in PercentileMain
• Use RetryDataFileIdOpenFile to prevent the flags from being modified.
• Set compressByteConvert/compressDiffConvert to true when defElem is set to true.
• Bugfix: mdtruncate redo can not read compress_opt.
• Bugfix: backup failed when uncompressed page in compressed table.
• Fixed the bug of join (like...including reloptions) will success even if reloptions contains compresstype compresstype can be used with (semgent=off) reloption.
• Bugfix: poor index compression ratio when use default pglz compression top standby repair.
• Bugfix: poor index compression ratio when use default pglz compression.
• change fastcheck.
• Fix partition error.
• Fix record type issue caused by composite datum.
• Fix create index info.
• Fix generated issue.
• Fix the core problem caused by undetermined return value during hash_index playback.
• Fix the issue of man-in-the-middle attack in SSL authentication mode.
• Fix the lower bound integer rounding problem of array_set function.
• gs_ctl adds -C parameter to adapt to churn tolerance.
• Repair opengauss close failed issue.
• Repair GSC no cache issue.
• Repair gs_gsc_memory_detail memory info inaccurate issue.
• Repair wal_level=archive in switchover/failover get tuple concurrently updated issue.
• Repair gsc sub thread exit error issue.
• Solve the switchover problem of pagewriter exiting.
• Solve the switchover bugs caused by the wake_sub_threads.
• [Round Main] fix bug interval partition exchange failure.
• Fix the bug that the pgplsql invokes the syntax parser error.
• Fix publish subscription deadlock issue.
• Fix the bug that the DefElem type in the logical.h file is not declared when the wal2json plugin is compiled.
• smp bugfix.
• Fix the coredump caused when the switching between the primary server and standby server is frequent.
• Change log of alter compresstype\compress_chunk_size.
• Optimize the message for create table.
• Reserve 160kb where backup compressed file.

5. Known Issues

No.	Problem Description	Impact Analysis	Evasion and Contingency Measures
1	The coredump problem probably occurred in the standby database under the primary/standby pressure test	Risk: high (impact level: high; probability of occurrence: low) When the primary database frequently creates database and drops database, it will cause the standby database to load the invalid data pages into the buffer pool when it plays back the data, thus causing a coredump.	Workaround: Avoiding frequent database creation and drop operations can reduce the probability of occurrence, but it cannot be completely avoided. Contingency measure: N/A
2	In the primary/standby pressure test, the standby database may have the problem that some modified data cannot be found.	Risk: high (impact level: high; probability of occurrence: medium) In some cases, querying data from the standby database will not find the data that has just been updated, and the phenomenon is similar to data being deleted.	Workaround: When the standby database is queried, set the hot_standby_feedback parameter to avoid the data that needs to be queried from being reclaimed prematurely by the primary database. Contingency measure: N/A
3	After the version upgrade, operators such as >, >=, <, <=, =, etc. can be used for point types. In general, the point type cannot use these operators.	Risk: low (impact level: low; probability of occurrence: high) After the upgrade, the operators that correspond to the point type works, but an error is still reported.	Workaround: This type is not an ordinary data type and is not usually used. If you do need to report an error, you can manually remove the operator metadata corresponding to the type to solve it. Contingency measure: N/A
4	After the upgrade, the Oracle dbms_job.user_export can return results when the parameters are illegal. In general, an illegal parameter error is reported.	Risk: low (impact level: low; probability of occurrence: high) After the upgrade, the arm machine can return results when the parameter is illegal.	Workaround: It is considered normal. If you do need to report an error here, you need to reinstall the Oracle compatibility extension. Contingency measure: N/A