3.4.1 (10 February 2016)

General

• Update license copyright to 2016
• Use slim python version with Docker #527 (xaka)
• Changed --master-only exit code to 0 when connected to non-master node #540 (wkruse)
• Add cx_Freeze capability to setup.py, plus a binary_release.py
  script to simplify binary package creation. #554 (untergeek)
• Set Elastic as author. #555 (untergeek)
• Put repository creation methods into API and document them. Requested in #550 (untergeek)

Bug fixes

• Fix sphinx documentation build error #506 (hydrapolic)
• Ensure snapshots are found before iterating #507 (garyelephant)
• Fix a doc inconsistency #509 (pmoust)
• Fix a typo in show documentation #513 (pbamba)
• Default to trying the cluster state for checking whether indices are closed, and
  then fall back to using the _cat API (for Amazon ES instances). #519 (untergeek)
• Improve logging to show time delay between optimize runs, if selected. #525 (untergeek)
• Allow elasticsearch-py module versions through 2.3.0 (a presumption at this point) #524 (untergeek)
• Improve logging in snapshot api method to reveal when a repository appears to be
  missing. Reported in #551 (untergeek)
• Test that --timestring has the correct variable for --time-unit.
  Reported in #544 (untergeek)
• Allocation will exit with exit_code 0 now when there are no indices to work on.
  Reported in #531 (untergeek)

3.4.0 (28 October 2015)

General

• API change in elasticsearch-py 1.7.0 prevented alias operations. Fixed in
  #486 (HonzaKral)
• During index selection you can now select only closed indices with --closed-only.
  Does not impact --all-indices Reported in #476. Fixed in #487 (Basster)
• API Changes in Elasticsearch 2.0.0 required some refactoring. All tests pass
  for ES versions 1.0.3 through 2.0.0-rc1. Fixed in #488 (untergeek)
• es_repo_mgr now has access to the same SSL options from #462. #489 (untergeek)
• Logging improvements requested in #475. (untergeek)
• Added --quiet flag. #494 (untergeek)
• Fixed index_closed to work with AWS Elasticsearch. #499 (univerio)
• Acceptable versions of Elasticsearch-py module are 1.8.0 up to 2.1.0 (untergeek)

3.4.0.rc1 (23 October 2015)

Announcement

This release supports Elasticsearch 2.0. The CI Jenkins build tests with 2.0 are currently not working because of Jenkins issues rather than Curator code. All tests pass with python 2.7 and 3.4 with my hardware.

General

• API change in elasticsearch-py 1.7.0 prevented alias operations. Fixed in
  #486 (HonzaKral)
• During index selection you can now select only closed indices with --closed-only.
  Does not impact --all-indices Reported in #476. Fixed in #487 (Basster)
• API Changes in Elasticsearch 2.0.0 required some refactoring. All tests pass
  for ES versions 1.0.3 through 2.0.0-rc1. Fixed in #488 (untergeek)
• es_repo_mgr now has access to the same SSL options from #462. #489 (untergeek)
• Logging improvements requested in #475. (untergeek)
• Added --quiet flag. #494 (untergeek)
• Fixed index_closed to work with AWS Elasticsearch. #499 (univerio)

3.3.0 (31 August 2015)

New Features

• Allocation type can now also be include or exclude, in addition to the
  the existing default require type. Add --type to the allocation command
  to specify the type. #443 (steffo)
• Add SSL features, --ssl-no-validate and certificate to provide other
  ways to validate SSL connections to Elasticsearch. #436 (untergeek)

General

• Bump elasticsearch python module dependency to 1.6.0+ to enable synced_flush
  API call. Reported in #447 (untergeek)

Bug fixes

• Delete by space was only reporting space used by primary shards. Fixed to
  show all space consumed. Reported in #455 (untergeek)
• Update exit codes and messages for snapshot selection. Reported in #452 (untergeek)
• Fix potential int/float casting issues. Reported in #465 (untergeek)

Testing

• Curator is tested in Jenkins. Each commit to the master branch is tested with both Python versions 2.7.6 and 3.4.0 against each of the following Elasticsearch versions:
  • 1.7_nightly
  • 1.6_nightly
  • 1.7.0
  • 1.6.1
  • 1.5.1
  • 1.4.4
  • 1.3.9
  • 1.2.4
  • 1.1.2
  • 1.0.3
• Jenkins was testing with Python 2.6.3 and 3.2.1. Those Python versions were buggy, so they've been updated to the aforementioned 2.7.6 and 3.4.0. If you are using a Python version different from either of these, your results may vary.

3.2.3 (16 July 2015)

Bug fix

• In order to address customer and community issues with bulk deletes, the
  master_timeout is now invoked for delete operations. This should address
  503s with 30s timeouts in the debug log, even when --timeout is set to
  a much higher value. The master_timeout is tied to the --timeout
  flag value, but will not exceed 300 seconds. #420 (untergeek)

General

• Mixing it up a bit here by putting General second! The only other changes
  are that logging has been improved for deletes so you won't need to have the
  --debug flag to see if you have error codes >= 400, and some code
  documentation improvements.

3.2.2 (13 July 2015)

General

• This is a very minor change. The mock library recently removed support
  for Python 2.6. As many Curator users are using RHEL/CentOS 6, which is
  pinned to Python 2.6, this requires the mock version referenced by Curator
  to also be pinned to a supported version (mock==1.0.1).

Also note that build failures were related to this. Mock 1.1.0 and 1.1.1 depend on pbr, which was at 1.2.0 when Curator 3.2.1 was released. The pbr module has since been upgraded to 1.3.0, which likely fixed the errors. Since we've pinned mock, this is moot, however.

3.2.1 (10 July 2015)

This is a lightweight patch release.

General

• Added delete verification & retry (fixed at 3x) to potentially cover an edge
  case in #420 (untergeek)
• Since GitHub allows rST (reStructuredText) README documents, and that's what
  PyPI wants also, the README has been rebuilt in rST. (untergeek)

Bug fixes

• If closing indices with ES 1.6+, and all indices are closed, ensure that the
  seal command does not try to seal all indices. Reported in #426 (untergeek)
• Capture AttributeError when sealing indices if a non-TransportError occurs.
  Reported in #429 (untergeek)

3.2.0 (25 June 2015)

3.2.0 (25 June 2015)

New!

• Added support to manually seal, or perform a synced flush
  on indices with the seal command. #394 (untergeek)
• Added experimental support for SSL certificate validation. In order for
  this to work, you must install the certifi python module:
  pip install certifi
  This feature should automatically work if the certifi module is
  installed. Please report any issues.

General

• Changed logging to go to stdout rather than stderr. Reopened #121 and
  figured they were right. This is better. (untergeek)
• Exit code 99 was unpopular. It has been removed. Reported in #371 and #391
  (untergeek)
• Add --skip-repo-validation flag for snapshots. Do not validate write
  access to repository on all cluster nodes before proceeding. Useful for
  shared filesystems where intermittent timeouts can affect validation, but
  won't likely affect snapshot success. Requested in #396 (untergeek)
• An alias no longer needs to be pre-existent in order to use the alias
  command. #317 (untergeek)
• es_repo_mgr now passes through upstream errors in the event a repository
  fails to be created. Requested in #405 (untergeek)

Bug fixes

• In rare cases, * wildcard would not expand. Replaced with _all.
  Reported in #399 (untergeek)
• Beginning with Elasticsearch 1.6, closed indices cannot have their replica
  count altered. Attempting to do so results in this error:
  org.elasticsearch.ElasticsearchIllegalArgumentException: Can't update [index.number_of_replicas] on closed indices [[test_index]] - can leave index in an unopenable state
  As a result, the change_replicas method has been updated to prune closed
  indices. This change will apply to all versions of Elasticsearch.
  Reported in #400 (untergeek)
• Fixed es_repo_mgr repository creation verification error. Reported in #389
  (untergeek)

3.1.0 (21 May 2015)

Announcement

This is a minor version release with a few new features and some bug fixes.

• Test count: 255
• Code coverage: 99%

General

• If wait_for_completion is true, snapshot success is now tested and logged.
  Reported in #253 (untergeek)
• Log & return false if a snapshot is already in progress (untergeek)
• Logs individual deletes per index, even though they happen in batch mode.
  Also log individual snapshot deletions. Reported in #372 (untergeek)
• Moved chunk_index_list from cli to api utils as it's now also used by filter.py
• Added a warning and 10 second timer countdown if you use --timestring to filter
  indices, but do not use --older-than or --newer-than in conjunction with it.
  This is to address #348, which behavior isn't a bug, but prevents accidental
  action against all of your time-series indices. The warning and timer are
  not displayed for show and --dry-run operations.
• Added tests for es_repo_mgr in #350
• Doc fixes

Bug fixes

• delete-by-space needed the same fix used for #245. Fixed in #353 (untergeek)
• Increase default client timeout for es_repo_mgr as node discovery and
  availability checks for S3 repositories can take a bit. Fixed in #352 (untergeek)
• If an index is closed, indicate in show and --dry-run output.
  Reported in #327. (untergeek)
• Fix issue where CLI parameters were not being passed to the es_repo_mgr
  create sub-command.
  Reported in #337. (feltnerm)

3.0.3 (27 Mar 2015)

Announcement

This is a bug fix release. #319 and #320 are affecting a few users, so this
release is being expedited.

• Test count: 228
• Code coverage: 99%

General

• Documentation for the CLI converted to Asciidoc and moved to
  http://www.elastic.co/guide/en/elasticsearch/client/curator/current/index.html
• Improved logging, and refactored a few methods to help with this.
• Dry-run output is now more like v2, with the index or snapshot in the log
  line, along with the command. Several tests needed refactoring with this
  change, along with a bit of documentation.

Bug fixes

• Fix links to repository in setup.py. Reported in #318 (untergeek)
• No more --delay with optimized indices. Reported in #319 (untergeek)
• --request_timeout not working as expected. Reinstate the version 2
  timeout override feature to prevent default timeouts for optimize and
  snapshot operations. Reported in #320 (untergeek)
• Reduce index count to 200 for
  test.integration.test_cli_commands.TestCLISnapshot.test_cli_snapshot_huge_list
  in order to reduce or eliminate Jenkins CI test timeouts.
  Reported in #324 (untergeek)
• --dry-run no longer calls show, but will show output in the log, as
  in v2. This was a recurring complaint. See #328 (untergeek)