.. _version_2.3.4:

=============
Version 2.3.4
=============

Released on 2018/03/06.

If you are upgrading a cluster, you must be running CrateDB
:ref:`version_1.1.3` or higher before you upgrade to 2.3.4.

If you want to perform a `rolling upgrade`_, your current CrateDB
version number must be at least :ref:`version_2.3.0`. Any upgrade
from a version prior to this will require a `full restart upgrade`_.

.. WARNING::

   Before upgrading, you should `back up your data`_.

.. _rolling upgrade: http://crate.io/docs/crate/guide/best_practices/rolling_upgrade.html
.. _full restart upgrade: http://crate.io/docs/crate/guide/best_practices/full_restart_upgrade.html
.. _back up your data: https://crate.io/a/backing-up-and-restoring-crate/

Changelog
=========

Changes
-------

- Updated Elasticsearch to v5.6.8.

Fixes
-----

- Fixed the URL linking to stats collection documentation in the monitoring
  tab of the Admin UI.

- Updated the table list search field to filter tables only by table name and
  table schema in the Admin UI.

- Fixed an issue in the Admin UI that caused the table list to display wrong
  results.

- Re-enable implicit casting for columns whenever columns are used on both
  sides of an operator or function. The implicit cast will obey the type
  precedence (eg. long > integer).

- Fixed an issue that caused incorrect results to be returned when a ``WHERE``
  clause filters on a partitioned and an non-partitioned column. E.g.::

    SELECT * FROM t WHERE parted_col='value' AND other_col='some_value'

- Fixed an issue that causes ``JOINS`` with certain ``ORDER BY`` constructs to
  fail.

- Fixed an issue where a ``NullPointerException`` would be thrown when trying
  to reroute shards to and from non-data nodes using ``ALTER TABLE REROUTE``.

- Fixed an issue that would cause queries and data manipulation statements with
  explicit ``WHERE`` conditions to throw ``NullPointerException`` when running
  against a table with a composite primary key.

- Fixed a race condition that could cause the decommissioning of nodes using
  the graceful shutdown procedure to interrupt the execution of active queries.

- Read only queries should no longer fail with a ``TableUnknownException`` or
  ``IndexNotFoundException`` in case of a shard relocation.
