This topic describes the enhancements and bug fixes in Hologres V4.1.17.
Enhancements
-
Enhanced retention metrics for adaptive query execution: Adds support for intermediate partial aggregation result types for range retention count metrics. This improves the stability and availability of the corresponding metrics pipeline.
-
Decoupled incremental broadcast penalty logic: The Query Optimizer (QO) now uses an optimized penalty strategy for incremental broadcasts, which reduces interference from unnecessary plan penalties.
-
Enhanced statistics for partition tables: Introduces partition Most Common Values (MCV) for partition tables to accelerate statistics collection for partition tables of all types.
Bug fixes
|
Severity |
Title |
Description |
Affected versions |
Workaround |
|
P2 |
Fixed an error in aggregate pushdown with left join. |
Fixed an optimization error that occurred during an aggregate pushdown below a left join. This issue could produce incorrect results. |
V4.1 and later |
Upgrade to V4.1.17 or later. |
|
P2 |
Fixed an error in incremental aggregation with all-NULL group keys. |
Fixed an issue where the incremental |
V4.1 and later |
|
|
P2 |
Fixed a core dump during compaction of v0.6.x ORC files with bitmaps. |
Fixed an issue that could cause a core dump during the compaction of v0.6.x ORC files containing bitmaps. |
V4.1 and later |
|
|
P2 |
Fixed a stack overflow risk during DXL translation. |
Added stack protection to the DXL translation pipeline to prevent crashes in deep translation scenarios. |
V4.1 and later |
|
|
P2 |
Fixed a core dump in the FE after a re-thrown exception. |
Fixed a core dump in the Frontend (FE) caused by re-throwing an exception after a |
V4.1 and later |
|
|
P1 |
Fixed the default behavior for endpoints without a specified protocol. |
Fixed incorrect default protocol handling for endpoints without an explicit URL scheme, which could cause access failures. |
V4.1 and later |
|
|
P2 |
Fixed incorrect matching by empty query queue classification rules. |
Fixed an issue that caused empty query queue classification rules to match incorrectly, which affected the results of query queue classification. |
V4.1 and later |
|
|
P2 |
Fixed the erroneous deletion of unpersisted snapshots. |
Fixed an issue that caused the erroneous deletion of unpersisted snapshots, which could compromise the stability of the snapshot process. |
V4.1 and later |
|
|
P2 |
Fixed percentile calculation for inputs with a very large number of rows. |
Fixed an issue where percentile calculations failed to reject input with more than |
V4.1 and later |