Skip to content

Releases: apache/airflow

Airflow 1.10.5, 2019-08-30

04 Sep 10:35
Compare
Choose a tag to compare

New Features

  • [AIRFLOW-1498] Add feature for users to add Google Analytics to Airflow UI (#5850)
  • [AIRFLOW-4074] Add option to add labels to Dataproc jobs (#5606)
  • [AIRFLOW-4846] Allow specification of an existing secret containing git credentials for init containers (#5475)

Improvements

Bug fixes

Misc/Internal

Doc-only changes

  • [AIRFLOW-XXX] Fixed Azkaban link (#5865)
  • [AIRFLOW-XXX] Remove duplicate lines from CONTRIBUTING.md (#5830)
  • [AIRFLOW-XXX] Fix incorrect docstring parameter in SchedulerJob (#5729)

1.10.4rc3

17 Jul 15:09
Compare
Choose a tag to compare
1.10.4rc3 Pre-release
Pre-release
[AIRFLOW-XXX] Update changelog for 1.10.4rc3

1.10.3

17 Jul 15:10
Compare
Choose a tag to compare

New Feature

Improvement

Bug fixes

Read more

1.7.0

28 Mar 19:32
Compare
Choose a tag to compare

Raw changelog

for the record, generated with:
~/node_modules/github-changes/bin/index.js -o airbnb -r airflow --only-pulls --use-commit-body --between-tags 1.6.2...1.7.0rc1 --token {your_gh_token_here}

1.7.0

Read more

1.6.2

04 Jan 20:29
Compare
Choose a tag to compare

1.6.2 is mostly a release with bug fixes and a few relatively minor features. Thanks to all contributors!

v1.6.1

15 Nov 06:20
Compare
Choose a tag to compare
  • Scheduler bugfix

v1.6.0

13 Nov 01:10
Compare
Choose a tag to compare

v1.6.0 brings:

  • [scheduler] the notion of DAG runs allows for more parallelization, and controls around scheduling (max number of running task instance per DAG, max number of DAG runs to be evaluated for scheduling, ...)
  • [scheduler] support for "externally triggered" DAG runs, or DAGs that run on demand as opposed to on a schedule
  • [scheduler] support for cron-like syntax (as in: "0 0 * * *") and macros (as in "@montlhly", "@hourly", "@Weekly", ...)
  • UI changes related to new scheduler features
  • LDAP authentication for the web UI, more extensible authentication backend
  • UI activity logging
  • WebHdfsSensor and Hook for HDFS interactions that are py3 compatible
  • Continuous integration with Travis-UI and Coveralls
  • ShortCircuitOperator
  • python3 compatibility!
  • Tons of bug fixes and incremental improvements
  • + all the things I'm forgetting while browsing through an infinite list of commits!

Thanks to everyone in the community for all the PRs (stellar contributions!), comments and issue reporting.

v1.5.2

23 Oct 06:13
Compare
Choose a tag to compare

This is probably missing some important items, but most of it should be captured

  • Initial setup on Travis CI provides continuous Integration, automated testing, no Hadoop unit tests just yet, but that is coming up
  • Unit tests coverage reports with coverals
  • Better py3 compatibility, unit tests run against both 2.7 ad 3.4 and we're now using the from __future__ imports to prevent regression
  • A MesosExecutor to run your tasks on Mesos
  • Some Kerberos integration for Hive / Hadoop
  • DAG's graph view legend for states are now toggles to highlight tasks in specific states
  • Automated zombie task instance killing as part of the scheduler's routine. The process looks for running tasks that don't have a heartbeat and kills them
  • MySqlHook bulk load option
  • More options in the UI's Mark Success form
  • MySQL uses mysqlclient lib instead of mysql-python
  • Using gunicorn instead of tornado as the wsgi web server
  • OracleHook
  • FTPHook
  • Much more! tons of bug fixes and usability improvements.

1.5.1

04 Sep 16:49
Compare
Choose a tag to compare

Bugfix around XCom table creation timestamp issue

1.5.0

04 Sep 06:29
Compare
Choose a tag to compare

v1.5.0 is a huge release. Tons of important features.

Make sure to run airflow upgradedb after you upgrade

Improvements:

  • @jlowin landed a feature to communicate information across tasks called XCom, more information here
  • @neovintage integrated Airflow with Alembic, making database migration easy, run airflow upgradedb to get your database up to date as you upgrade Airflow
  • The dependency engine is now more flexible, allowing for trigger rules, before this update, tasks would only be triggered when all dependencies were successful (still the default), now you can set tasks to trigger when a single parent succeeds, when one fails, when they all fail, or to fire regardless of their dependencies:
  • @neovintage added support for connection to be defined in environment variables, allowing to bypass or override the metadata database
  • @jlowin improved the tree view to render a non expendable when trees go above 5k nodes
  • The Druid hook and HiveToDruidTransfer are maturing and becoming production grade
  • @kapil-malik added a UI feature where some users can only see their own DAGs while superusers can still see all, this feature is turned off by default
  • Allowing all operators and some preoperators to receive list of sql statements to be execute as a chain
  • Improved the task_instance table to log the operator name and queued timestamp
  • Passwords in the metadata can now be encrypted
  • Improvements to the unit tests (speed + coverage)
  • Bugfixes + more