New
- Multipartitioned assets with one time dimension can now depend on earlier partitions of themselves.
define_asset_job
now accepts ahooks
argument.- Added support for
sqlalchemy==2.x
- [ui] The Runs page has been revamped with better filtering support
- [ui] The automaterialize policy page for SDA’s using the experimental AutomaterializePolicy feature now indicates time periods where no materializations happened due to no materialization conditions being met
- [dagster-k8s] The Dagster Helm chart now includes an
additionalInstanceConfig
key that allows you to supply additional configuration to the Dagster instance. - [dagster-aws] The
EcsRunLauncher
now uses a different task definition family for each job, instead of registering a new task definition revision each time a different job is launched. - [dagster-aws] The
EcsRunLauncher
now includes arun_ecs_tags
config key that lets you configure tags on the launched ECS task for each run.
Bugfixes
- When a sensor had yield statement and also returned a
SkipReason
, theSkipReason
would be ignored. This has been fixed. - [dagster-cloud] Fixed a bug in the docker user code launcher that was preventing code location containers from being properly cleaned up.
- Fixed an issue where the Dagster UI would sometimes raise a ``RuntimeError: dictionary changed size during iteration` exception while code servers were being reloaded.
- Fixed an issue where the Dagster daemon reloaded your code server every 60 seconds when using the new experimental
dagster code-server start
CLI, instead of only reloading your code when you initiate a reload from the Dagster UI. - Fixed a GraphQL error which would occur when loading the default config for jobs without config.
- [dagster-dbt] Fixed an error which would arise when trying to load assets from a DBT Cloud instance using the Pythonic-style resource.
Community Contributions
- Added the ability to specify metadata on asset jobs, by adding the
metadata
parameter todefine_asset_job
(Thanks Elliot2718!) - [dagster-databricks] Connected databricks stdout to local stdout, to be handled by the compute log manager (Thanks loerinczy!)
- [dagster-census] Fixed
poll_sync_run
to handle the “preparing” status from the Census API (Thanks ldnicolasmay!)
Experimental
@observable_source_asset
-decorated functions can now return aDataVersionsByPartition
to record versions for partitions.@dbt_assets
DbtCliTask
's created by invokingDbtCli.cli(...)
now have a method.is_successful()
, which returns a boolean representing whether the underlying CLI process executed the dbt command successfully.- Descriptions of assets generated by
@dbt_assets
can now be customized by overridingDbtManifest.node_info_to_description
. - IO Managers can now be configured on
@dbt_assets
.
Documentation
- New guide on using Dagster to manage machine learning pipelines
Dagster Cloud
- Added support for streaming upload of compute logs to Dagster Cloud
- The ECS agent now supports setting
server_ecs_tags
andrun_ecs_tags
that apply to each service or task created by the agent. See the docs for more information. - Fixed run filtering for calls to
instance.get_run_partition_data
in Dagster Cloud.