Release Notes for Beta - 20170112

January 12, 2017

Upgrade Notes

This release cannot be run concurrently with older beta releases. Please stop all nodes running older releases before restarting any node with this version.

SQL Language Changes

  • The ALTER TABLE VALIDATE CONSTRAINT statement can now be used to validate foreign keys in the UNVALIDATED state. #12682
  • ALTER TABLE DROP COLUMN can now drop columns that are involved in indexes. If the index includes columns other than the one being dropped, the CASCADE modifier is required. #12782
  • Leading zeros in numeric literals are now ignored instead of causing the number to be parsed as octal. #12803
  • Collated strings can now be cast to the non-collated string type. #12820
  • Arithmetic with aggregated functions no longer crashes when the input table is empty. #12733
  • The SHOW COLUMNS statement now works for non-root users with the appropriate permissions. #12776
  • The information_schema.statistics.NON_UNIQUE field is now set correctly. #12793
  • It is now possible to use EXPLAIN statements as an input to other queries using a new square bracket syntax: SELECT MAX(level) FROM [EXPLAIN SELECT * FROM kv ORDER BY v]. #12637
  • The format of SQL error messages has been standardized. #12401

Command-Line Interface Changes

  • The prototype cockroach backup and cockroach restore commands have been removed (backup and restore functionality will be provided through SQL statements instead). #12788

Admin UI Changes

  • The set of graphs displayed in the UI has been changed. #12797
  • The node list and cluster pages has been updated. #12643 #12719
  • A new metric raftlog.behind is now reported to track how far behind raft processing is on a node. #12800

Bug Fixes

  • Ranges can no longer be rebalanced away from their current lease holder. #12598
  • Eliminated a source of incorrect “not lease holder” errors. #12542
  • Reduced spammy “purgatory” log messages. #12528

Performance Improvements

  • The maximum allowable clock offset has been increased from 250ms to 500ms. #12821
  • Replica destruction is now much faster. #12745
  • Raft snapshots and replica destruction no longer run concurrently. #12737
  • Raft snapshots are now managed by a new internal queue. #12686
  • Raft logs are now garbage collected more aggressively. #12686
  • Replicas that are behind are no longer considered targets for lease rebalancing. #12736
  • Tuned some RocksDB parameters. #12677

Doc Updates


This release includes 60 merged PRs by 22 authors. We would like to thank the following contributors from the CockroachDB community:

  • songhao
  • ghao