All notable changes to this project will be documented in this file.
- Support upgrades from 0.5.x charts.
- Support airflow 1.10.1 environment variables.
- Fixed prometheus healthchecks.
- Fixed a deployment upgrade issue.
- Fixed a migration issue in houston when upgrading.
- Added support for platform and airflow deployment alerts on prometheus data.
- Added support for injecting airflow environment variables at runtime through the UI.
- Added support for different airflow executors.
- Added user controls for dynamically adjusting resource allocation and constrains.
- Added airflow chart upgrade functionality.
- Added lots of new grafana dashboards for persistent storage, prometheus, registry, elastic, fluentd, airflow container state and more.
- Updated prometheus chart to fix healthcheck.
- Initial airflow 1.10.1 support.
- Bug fixes
- Support for Service Accounts and deploying from CI/CD pipelines
- Fixed ingress redirect issue
- Updated to Alpine 3.8 for all images including Airflow
- New automated build pipeline
- Support for upgrading individual airflow deployments to latest chart.
- Fix for airflow task logs link on taskinstances page.
- Support for the KubernetesPodOperator. Every deployment goes into its own namespace with resource quotas, limits and defaults.
- Support node auto-scaling. All containers are now assigned resource requests/limits.
- Support for database connection pooling. pgbouncer is now deployed with every deployment to pool connections to postgres for all airflow pods (webserver/scheduler/workers/kubepodoperators/etc)
- Grafana enhancements. New pgbouncer and nginx dashboards, fixed some queries (smooth out container metric graphs), and added deployment dropdowns to all dashboards to drill down into specific deployments.
- Fixed worker logs garbage collector
- Faster deploys. Fixed containers from restarting on initial boot, especially noticeable with flower.
- Larger default volume size for workers/prometheus/registry.
- Updated NGINX ingress controller.
- Some new options and configurations in the airflow chart (choosing executor, horizontal pod autoscaling, connection pooling, etc) that will make their way into the deployment configuration in the future.
- Other minor bug fixes and backend enhancements
- Added local email/password authentication option
- Integration with Auth0
- Default installation can now use local, Google, or Github as auth providers
- Fixed some Airflow webserver stability issues with a better healthcheck
- Fixed Grafana healthcheck
- Updated Grafana to v5.2.2
- Removed unused Prometheus target
- Larger default volume size on airflow workers
- Prevent redis and flower from restarting during deployments.
- All charts protected with NetworkPolicies.
- Improved docker caching, resulting in faster development and faster deployments.
- Added nginx-auth secret to houston, fixing issue with airflow deploys being created unsecured
- Set astronomer to be the only chart to install by default
- Updated houston chart to save global values as an env
- Updated commander chart to have elevated tiller permissions
- Updated all chart to include version in the chart label
- Upgrade setuptools in Airflow image
- Replace Phoenix with Houston
- Add /healthz to default backend
- Add platform default backend
- Initial public release
- Add relabelling rules to prometheus and updated metric names in grafana dashboard.
- Added nginx image.
- Removed basic auth via env from registry in favor of nginx.
- Bump commander and phoenix. Adds support for private docker registry.
- Entrypoint script for docker registry.
- Added Module/Component labels.
- Fixed phoenix image.
- Added Registry.
- Added Commander.
- Added Phoenix.
- Added labels for open.
- Changed Airflow worker-gc environment variable name.
- Added support for clearing worker logs after x days.
- Removed Snakebite python package due to compatibility issues with python 3.
- Fixed busted grafana build.
- Added support for passing default prometheus host for grafana datasource.
- Changed Prometheus config file name and paths.
- Added Prometheus Marathon config.
- Added Python and pip symlinks to python3 and pip3.
- Removed usage of .astro directory in start script.
- Added vendor images.
- Added Tini init system.
- Added support for Airflow onbuild tags.
- Added this Changelog
- Added StatsD package to Airflow
- Added StatsD exporter to stack.
- Added Prometheus to stack.
- Added Grafana to stack.
- Airflow scripts now run docker-compose with -d flag.