Version 6.4.5¶

Released on 2026-09-16.

Note

If you are upgrading a cluster, you must be running CrateDB 5.0.0 or higher before you upgrade to 6.4.5.

We recommend that you upgrade to the latest 6.3 release before moving to 6.4.5.

A rolling upgrade from >= 6.3.0 to 6.4.5 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 ¶

HTTP clients sending an invalid JWT token in the Authentication header now receive a 401 error instead of 500.
Fixed an issue that could cause statements including a cross join with two other relations to apply an incorrect filter if one of the join conditions included an equijoin where one relation appeared multiple times on one side of the equal condition. An example query:
```
SELECT * FROM t1 CROSS JOIN t2 INNER JOIN t3 ON (t1.x + t1.x) = t3.z AND t3.z = t2.y
```
Fixed an issue that could cause statements including a cross join and an outer join plus any other join to drop the outer join condition, leading to incorrect results.
Fixed an issue that caused casting string values to a longer bit string to fail with an StringIndexOutOfBoundsException instead of filling up the gap with 0.
Fixed an issue that caused casting null to the bit type to fail.
Fixed an issue that caused CrateDB to choose the wrong relation if a parent relation alias clashed with a local scoped relation name.
Fixed decoding of interval values sent from PostgreSQL wire protocol clients using text format in the x hours y minutes ... style.
Fixed an issue that could lead to out of memory errors if using the percentile aggregation.
Fixed an issue causing cache to retain some heavy structures, potentially leading to an OutOfMemoryError.

Fixed an issue where a select statement containing a correlated subquery and a group by would lose the subquery and result in an error, e.g.:

SELECT o.c1,
   (SELECT i.c1
    FROM source_symbols_t AS i
    WHERE i.c1 > o.c1
    ORDER BY i.c1
    LIMIT 1) AS next_c1
FROM source_symbols_t AS o
GROUP BY o.c1
ORDER BY o.c1;

> Couldn't create execution plan from logical plan because of:
Couldn't find (SELECT c1 FROM (i)) in SourceSymbols

Fixed performance regression introduced in Version 6.3.0 that caused CREATE SNAPSHOT to run slower if repository was created with Azure storage.
Fixed an issue that could cause an error message to reference a wrong relation if a column used in a query could not be resolved and multiple relations are involved. This could also lead to privileges being checked on the wrong relation while generating the error.
Fixed a regression introduced in Version 6.4.0 which caused a join, executed with the hash join implementation, to not match rows when the join condition used CHAR(n) join keys of different length but same value, because of the blank padding, which is applied to match the defined n length.

Version 6.4.5¶

Fixes¶

Fixes ¶