New Features and Resolved Issues in 8.3.0.x
8.3.0.110
Category |
Feature or Resolved Issue |
Cause |
Version |
Handling Method |
---|---|---|---|---|
New features |
By default, the global deadlock detection function is enabled for newly installed clusters. In earlier versions, this is disabled by default. |
- |
- |
- |
Resolved issues |
When a common table expression (CTE) is referenced only by another CTE for multiple times, null pointer access may be triggered, causing a fault. |
When a CTE is referenced multiple times exclusively by another CTE, it can result in incorrect recursive traversal logic. This may trigger a null pointer access and cause a fault in the cluster. |
8.3.0.101 |
Upgrade the version to 8.3.0.110 or later. |
If a query statement for a column-store table contains multiple result set functions, the error message " Set-returning function is not supported in vector engine" is displayed when the statement is executed. |
The vectorized executor on the non-table scan operator does not support multi-result-set functions. |
8.3.0.101 |
||
When the cluster is upgraded to 8.3.0, the thread pool function becomes invalid. |
When upgrading the cluster to version 8.3.0, adjust the max_stream_pool setting based on the current hardware configuration. Note that max_stream_pool is set to 0 due to an abnormal default unit identification. |
8.3.0.101 |
||
The error message "The parameter destMax is too small or parameter count is larger than macro parameter" is displayed when a Parquet foreign table is used for query. |
The buffer space for string-type fields in the Parquet foreign table is not properly managed, causing insufficient memory allocation for fields larger than 1 KB. This, in turn, leads to service errors. |
8.3.0.101 |
||
When performing a partition exchange operation on redistributed cold and hot tables, the error message "Multi-temperature table only supports column partition table" is displayed. |
The temporary table cannot be created during redistribution because it references the attributes of the original cold and hot tables. |
8.3.0.101 |
8.3.0.108
Category |
Feature or Resolved Issue |
Cause |
Version |
Handling Method |
---|---|---|---|---|
New features |
None |
- |
- |
- |
Resolved issues |
Lock Timeout Occurs During Concurrent Copy to the Database |
The lock API is being misused in the release of the lock by RecoverFromDelta, causing it to remain locked. |
8.3.0.105 |
Upgrade the version to 8.3.0.108 or later. |
8.3.0.105
Category |
Feature or Resolved Issue |
Cause |
Version |
Handling Method |
---|---|---|---|---|
New features |
None |
- |
- |
- |
Resolved issues |
A result set error can occur during multiple count distinct operations with SHARE SCAN. |
This issue arises from the incorrect use of upper-layer equivalence class information in the CTE subquery, which results in misidentifying the distribution column and omitting the necessary data redistribution operator. |
Versions earlier than 8.3.0.105 |
Upgrade the version to 8.3.0.105 or later. |
System catalog autovacuum is not functioning properly after upgrading to version 830. |
Prior to the upgrade, a DR migration was carried out, which involved setting autovacuum_max_workers to 0 and then changing it to the default value of 3. However, the value was not set to the expected value of 6 after the upgrade. |
Versions earlier than 8.3.0.105 |
||
If a UDF receives data in recv mode, an error might not be captured even if error is EAGAIN after a timeout. |
This occurs when cn_retry is enabled and the socket recv times out for 20 seconds, triggering error reporting logic instead of capturing the EAGAIN error code. |
Versions earlier than 8.3.0.105 |
8.3.0.103
Type |
Feature or Resolved Issue |
Cause |
Version |
Handling Method |
---|---|---|---|---|
New features |
json/jsonb supports vectorization. |
- |
- |
- |
Resolved issues |
Core dumps occur occasionally with ORC foreign tables. |
During foreign scan query pushdowns on ORC foreign tables, if the query constraint is a Boolean expression with implicit conversion and uses the is null condition, abnormal internal type conversion and memory overwriting may occur (RelabelType is forcibly converted to var)(RelabelType is forcibly converted to var). |
Versions earlier than 8.3.0.103 |
Upgrade the version to 8.3.0.103 or later. |
Cstore buffer error during 10 million batch upserts. |
The cstore buffer's cache elimination mechanism is Clock Swap, which operates a maximum of three cycles every three seconds. This limitation prevents the elimination of a large number of CUs (Usage Count > 3). |
Versions earlier than 8.3.0.103 |
||
Delta table space not released after import tasks stop. |
Deleted records in the hstore opt table may not be cleared, preventing space from being released. |
Versions earlier than 8.3.0.103 |
8.3.0.101
Type |
Feature or Resolved Issue |
Cause |
Version |
Handling Method |
---|---|---|---|---|
New features |
VACUUM FULL adapts to binlog tables. |
- |
- |
- |
Resolved issues |
Executing the analyze statement results in an error message stating "Could not open file 'pg_clog/000000000075'." |
Performing VACUUM FULL on a column-store table can prematurely reclaim clog files, causing access issues during ANALYZE after an active/standby switchover. |
8.3.0.100 |
Upgrade the version to 8.3.0.101 or later. |
Error "duplicate key value violates unique constraint 'pg_jobs_id_index', Key(job_id)=() already exist" during high concurrency job creation with \parallel on. |
The system failed to acquire the necessary lock for assigning unique job_ids. Consequently, concurrent processes attempting to obtain the same job_id led to error reports. |
8.3.0.100 |
||
Redistribution fails due to inability to skip specified tables. |
In multiple node group scenarios, redistribution cannot skip specified tables. |
8.3.0.100 |
||
Hstore test case failure due to faulty network adapter. |
Declaring a temporary variable without assigning a value leads to unexpected parameter values and excessively negative memory usage, causing test case execution failure when the network adapter is faulty. |
8.3.0.100 |
Feedback
Was this page helpful?
Provide feedbackThank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.
For any further questions, feel free to contact us through the chatbot.
Chatbot