July 12, 2021
Get future release notes emailed to you:
Downloads
Warning:
The CockroachDB executable for Windows is experimental and not suitable for production deployments. Windows 8 or higher is required.
Docker image
$ docker pull cockroachdb/cockroach:v20.2.13
Operational changes
- Added logs for important events during the server draining/shutdown process. #66882 Specifically, the following:
- Log when the server closes an existing connection while draining
- Log when the server rejects a new connection while draining
- Log when the server cancels in-flight queries after waiting for the duration of the
server.shutdown.query_waitcluster setting to elapse while draining.
DB Console changes
- Fixed an issue with displaying more than 100 hours of remaining time on the Jobs page. #66704
- Removed width styling on the multi bar. #66737 cockroachdb/cockroach#66737
- Fixed the Custom Chart debug page layout. #66623
- Added a drag-to-select timescale feature to the Custom Chart debug page. #66623
- Fixed a styling issue with a filter component on the CockroachCloud console. #66566
- Removed an unnecessary call to
console log. #66566 cockroachdb/cockroach#66566 - Bumped the
cluster-uiversion to v20.2.1. #66566 - Fixed a bug where metrics pages in the DB Console would lose their scroll position on chart data updates. #67088
Bug fixes
- Fixed a bug which prevented the optimizer from producing plans with partial indexes when executing some prepared statements that contained placeholders, stable functions, or casts. This bug was present since partial indexes were added in v20.2.0. #66641
- Fixed a panic that could occur in the optimizer when executing a prepared plan with placeholders. This could happen when one of the tables used by the query had computed columns or a partial index. #66832
- Fixed a bug that caused graceful drain to call
time.sleepmultiple times, which cut into the time needed for range lease transfers. #66852 - CockroachDB now allows a node with lease preferences to drain gracefully. #66714
- CockroachDB now avoids interacting with decommissioned nodes during DistSQL planning and consistency checking. #66951
- Changefeeds no longer interact poorly with large, abandoned transactions. It was previously possible for this combination to result in a cascade of work during transaction cleanup that could starve foreground traffic. #66814
- Changefeeds now properly invalidate cached range descriptors and retry when encountering decommissioned nodes. #67024
Performance improvements
- Added the ability to continue generating histograms when table statistics collection reaches memory limits, instead of disabling histogram generation. #67059
Contributors
This release includes 19 merged PRs by 16 authors.
Yes
No