Version 6.4.6 - Unreleased¶
Note
In development. 6.4.6 isn’t released yet. These are the release notes for the upcoming release.
Note
If you are upgrading a cluster, you must be running CrateDB 5.0.0 or higher before you upgrade to 6.4.6.
We recommend that you upgrade to the latest 6.3 release before moving to 6.4.6.
A rolling upgrade from >= 6.3.0 to 6.4.6 is supported. Before upgrading, you should back up your data.
Warning
Tables that were created before CrateDB 5.x will not function with 6.x and must be recreated before moving to 6.x.x.
You can recreate tables using COPY TO and COPY FROM or by
inserting the data into a new table.
Table of contents
See the Version 6.4.0 release notes for a full list of changes in the 6.4 series.
Fixes¶
Fixed a rounding error if casting timestamp or long values before 1970-01-01 to
date.Fixed an issue that caused order-sensitive aggregations (
string_agg,array_agg) applied on top ofUNION ALL ... ORDER BYto return incorrect results.Fixed an issue that could cause window functions using
RANGEwith integralPRECEDING/FOLLOWINGbounds to compute frames incorrectly due to boundary overflow.Fixed an issue that caused
CREATE SNAPSHOTto fail if repository had at least one snapshot created before and backend storage’s DELETE API was temporarily unavailable.Fixed an issue that caused
NULLto always be returned fornumeric_scaleof information_schema.columns table. Now it returns the definedscaleforNUMERIC,0for integral numeric types (BYTE,SHORT,INTEGER,LONG) andNULLfor all other types.Fixed an issue, introduced with Version 6.3.0, which caused RESTORE SNAPSHOT to always restore all foreign tables contained in the snapshot, even if they weren’t requested. Foreign tables are now only restored together with their servers, e.g. by using
RESTORE SNAPSHOT ... ALLorRESTORE SNAPSHOT ... METADATA. A foreign table restored explicitly withRESTORE SNAPSHOT ... TABLEdoesn’t restore its server and user mappings, so it’s unusable unless they already exist or created manually.Fixed a rare race condition causing the
queryCircuitBreaker’s state not to be reset afterKILL, and leading to consequent queries to fail with false positiveCircuitBreakingException.Fixed a race condition in
EXPLAIN ANALYZEthat could cause an intermittentNullPointerException.Fixed an issue that could cause logical replication of new partitions to fail after restarting publisher nodes.