dagster

Bugfixes

• [dagster-embedded-elt] Fixed an issue where EnvVars used in Sling source and target configuration would not work properly in some circumstances.
• [dagster-insights] Reworked the Snowflake insights ingestion pipeline to improve performance and increase observability.

New

• [ui] Asset graph now displays active filters.
• [ui] Asset graph can now be filtered by compute kind.
• [ui] When backfilling failed and missing partitions of assets, a “Preview” button allows you to see which ranges will be materialized.
• [dagster-dbt] When running DAGSTER_DBT_PARSE_PROJECT_ON_LOAD=1 dagster dev in a new scaffolded project from dagster-dbt project scaffold, dbt logs from creating dbt artifacts to loading the project are now silenced.
• [dagster-airbyte] Added a new connection_meta_to_group_fn argument which allows configuring loaded asset groups based on the connection’s metadata dict.
• [dagster-k8s] Debug information about failed run workers in errors surfaced by run monitoring now includes logs from sidecar containers, not just the main dagster container.

Bugfixes

• The QueuedRunCoordinatorDaemon has been refactored to paginate over runs when applying priority sort and tag concurrency limits. Previously, it loaded all runs into memory causing large memory spikes when many runs were enqueued.
• Callable objects can once again be used to back sensor definitions.
• UPathIOManager has been updated to use the correct path delimiter when interacting with cloud storages from a Windows process.
• In the default multiprocess executor, the STEP_WORKER_STARTED event now fires before importing code in line with the other executors.
• During execution, skipping a step now takes precedence over “abandoning” it due to upstream failure. This is expected to substantially improve the “retry from failure” workflow when conditional branching is in use.
• Fixed an issue where default config values set to EnvVar did not work properly.
• Fixed an issue where resources which implemented IAttachDifferentObjectToOpContext would pass the incorrect object to schedules and sensors.
• Fixed a bug that caused auto-materialize failures when using the materialize_on_cron rule with dynamically partitioned assets.
• Fixed an issue where sensor ticks would sporadically fail with a StopIteration exception.
• [ui] For a job launchpad with a large number of tabs, the “Remove all” option was pushed offscreen. This has been fixed.
• [ui] The asset backfill page now correctly shows backfills that target only unpartitioned assets.
• [ui] Launching an asset job that was defined without_checks no longer fails by attempting to include the checks.
• [dagster-databricks] fix bug that caused crash when polling a submitted job that is still in the Databricks queue (due to concurrency limit).

Community Contributions

• Patched issue where the local compute log path exposed file content outside of the compute log base directory - thanks r1b!
• [dagster-databricks] Added ability to authenticate using an Azure service principal and fix minor bugs involving authenticating with a service principal while DATABRICKS_HOST is set. Thanks @zyd14!

Experimental

• [ui] Dark mode is now available via the User Settings dialog, currently in an experimental state. By default, the app will use a “legacy” theme, closely matching our current colors. A new light mode theme is also available.
• [ui] Asset graph group nodes can be collapsed/expanded by right clicking on the collapsed group node or the header of the expanded group node.
• [ui] Asset graph group nodes can be all collapsed or all expanded by right clicking anywhere on the graph and selecting the appropriate action.
• [ui] The tree view was removed from the asset graph.
• [pipes] PipesLambdaClient, an AWS Lambda pipes client has been added to dagster_aws.
• Fixed a performance regression introduced in the 1.5.10 release where auto-materializing multi-assets became slower.

Documentation

• [getting-started] Added an overview to the Getting Started section that explains the whats and whys of Dagster.
• [pipes] Added a guide for using the new PipesLambdaClient with Dagster Pipes.
• [getting-started] Simplified the Getting Started category. The following pages have been moved:
  • Understanding Dagster project files is now in Guides
  • Telemetry is now in About
• [guides] Fixed a broken link in the Airflow-to-Dagster concept mapping guide.
• [deployment] Cleaned up and updated the Executing with Celery OSS deployment guide.
• [general] Added two guides that were previously missing to the side navigation:
  • Utilizing SCIM provisioning (Deployment > Cloud > Authentication & users > SCIM provisioning)
  • Pandera (Integrations > Pandera)

Dagster Cloud

• When a Dagster Cloud agent starts up, it will now wait to display as Running on the Agents tab in the Dagster Cloud UI until it has launched all the code servers that it needs in order to serve requests.

New

• Added a new MetadataValue.job metadata type, which can be used to link to a Dagster job from other objects in the UI.
• [asset backfills] Previously, when partitions definitions were changed after backfill launch, the asset backfill page would be blank. Now, when partitions definitions are changed, the backfill page will display statuses by asset.
• [dagster-bigquery, dagster-duckdb, dagster-snowflake]. The BigQuery, DuckDB, and Snowflake I/O Managers will now determine the schema (dataset for BigQuery) in the following order of precedence: schema metadata set on the asset or op, I/O manager schema/ dataset configuration, key_prefix set on the asset. Previously, all methods for setting the schema/dataset were mutually exclusive, and setting more than one would raise an exception.
• [dagster-shell] Added option to exclude the shell command from logs.
• [dagster-dbt] When running DAGSTER_DBT_PARSE_PROJECT_ON_LOAD=1 dagster dev in a new scaffolded project from dagster-dbt project scaffold, dbt artifacts for loading the project are now created in a static target/ directory.

Bugfixes

• Problematic inheritance that was causing pydantic warnings to be emitted has been corrected.
• It's now possible to use the logger of ScheduleEvaluationContext when testing via build_schedule_context.
• The metadata from a Failure exception is now hoisted up to the failure that culminates when retry limits are exceeded.
• Fixed bug in which the second instance of an hour partition at a DST boundary would never be shown as “materialized” in certain UI views.
• Fixed an issue where backfilling an hourly partition that occurred during a fall Daylight Savings Time transition sometimes raised an error.
• [auto-materialize] Fix issue where assets which were skipped because required parent partitions did not exist would not be materialized once those partitions came into existence.
• [dagster ecs] The exit code of failed containers is now included in the failure message.
• [dagster pipes] The PipesK8sClient now correctly raises on failed containers.
• [dagster pipes] Using pipes within ops instead of assets no longer enforces problematic constraints.
• [helm] Added maxCatchupRuns and maxTickRetries configuration options for the scheduler in the Helm chart.
• [embedded-elt] Fixed crashes for non-unicode logs.
• [UI] Fixed an issue where the test sensor dialog for a sensor that targeted multiple jobs would claim that all of the runs were targeting the same job.
• [UI] Asset keys, job names, and other strings in Dagster UI no longer truncate unnecessarily in Firefox in some scenarios
• [UI] A larger “View prior events” button on the Asset > Partitions page makes it easier to see the historical materializations of a specific partition of an asset.
• [asset-checks, dbt] Fixed a bug that that caused asset checks to not execute when a run was not a subset. As part of the fix, the default dbt selection selection string will not be used for dbt runs, even when not in a subset. Instead we pass the explicit set of models and tests to execute, with DBT_INDIRECT_SELECTION=empty.
• [asset-checks] Fixed a bug that caused asset checks defined with @asset(check_specs=... to not cooperate with the key_prefix argument of the load_assets_from_modules method and it’s compatriots.
• [asset-checks] Fixed a bug that caused errors when launching a job from the UI that excluded asset checks.
• [asset-checks] Fixed a bug that caused UI errors when a check run was deleted.

Deprecations

• Marked the experimental Airbyte ingestion-as-code feature as deprecated, to be removed in a future release. We suggest users interested in managing their Airbyte connections in code use the Airbyte terraform provider.

Community Contributions

• define_asset_job now accepts an op_retry_policy argument, which specifies a default retry policies for all of the ops in the job. (thanks Eugenio Contreras!)
• Fix IOManager not being able to load assets with MultiPartitionsDefinition - thanks @cyberosa!
• [dagster-essentials] Three typo fixes in Lesson 8 - thanks Colton @cmpadden!

Experimental

• The observable_source_asset decorator now accepts a key argument.
• [dagster pipes] an implicit_materializations argument has been added to get_results and get_materialize_result to control whether an implicit materialization event is created or not.
• [embedded-elt] Added a new builder and SlingConnectionResource to allow reusing sources and targets interoperably.
• [UI] Updated the experimental concurrency limits configuration page to show per-op runtime info and control.
• [UI] The Auto-materialize history tab for each asset now only includes rows for evaluations where the result of evaluating the policy has changed. Previously, it would also show a row in the table representing periods of time where nothing changed.
• [asset-checks, dbt] build_dbt_asset_selection now also selects asset checks based on their underlying dbt tests. E.g. build_dbt_asset_selection([my_dbt_assets], dbt_select="tag:data_quality") will select the assets and checks for any models and tests tagged with ‘data_quality’.

Documentation

• Added information about EnvVar vs. os.getenv to the Environment variables documentation.
• Updates to the Asset selection syntax reference, including expanded examples for Python, the CLI, and the Dagster UI.
• Added Experimental tags to all Dagster Cloud Insights docs.
• Updated the Helm - Migrating a Dagster instance while upgrading guide to include a prerequisites section.

Dagster Cloud

• Branch deployments now use the same timeouts for starting and canceling runs that are set for their parent full deployment, instead of a fixed value of 10 minutes.
• [k8s agent] Setting labels on a code location will now apply those labels to the kubernetes deployment and service for that code location, rather than just applying them to the pod for that code location.
• Dagster Insights is now available by default to all Cloud users, enabling analysis of Dagster runtime metrics at the platform and definition level. This feature is still in experimental mode, a label that we expect to remove in Q1 2024.

New

• [ui] Enabled collapsing asset groups in the global asset view when the new experimental asset graph is turned on in User Settings
• [ui] The experimental asset graph sidebar now supports keyboard navigation via arrow keys
• [ui] You can now right click nodes in the asset graph to materialize them or filter the graph
• [ui] Jobs can now be searched by run ID
• [ui] You can now launch runs from the job actions menu
• [auto-materialize] A new AutoMaterializeRule.materialize_on_cron() rule makes it possible to create policies which materialize assets on a regular cadence.
• [auto-materialize] If a partition-mapping-related error occurs within the Asset Daemon, a more informative error message will be provided.
• [dagster-databricks] Extended the set of available config options to the Databricks step launcher - thanks @zyd14!

Bugfixes

• Fixed an issue where some schedules incorrectly skipped ticks during Daylight Savings Times transitions.
• Returning a SensorResult from a sensor no longer overwrites a cursor if it was set via the context.
• Fixed issue which could cause incorrect execution order when executing multi-assets with can_subset=True alongside assets which were upstream of some assets in the multi-asset, and downstream of others.
• Previously, when creating an HourlyPartitionsDefinition with a non-UTC timezone and the default format string (or any format string not including a UTC-offset), there was no way to disambiguate between the first and second instance of the repeated hour during a daylight saving time transition. Now, for the one hour per year in which this ambiguity exists, the partition key of the second instance of the hour will have the UTC offset automatically appended to it.
• [asset checks] Fixed a bug that caused an error when passing check_specs to AssetsDefinition.from_graph
• [dagster-dbt] Fixed a bug in dagster-dbt that caused some dbt tests to not be selected as asset checks.
• [dagster-dbt] Fixed an issue where multiple copies of the dbt manifest were held in memory when loading a dbt project as software-defined assets.
• The email_on_failure sensor called deprecated methods on the context. This has been fixed

Community Contributions

• [dagster-deltalake] Added Delta Lake support along with support for pandas and polars. Thanks Robert Pack @roeap!
• [dagster-graphql] Fixed the asset cursor format to use a normalized format - thanks @sisidra!
• [dagster-databricks] Extended the set of available config options to the Databricks step launcher - thanks @zyd14!

Experimental

• DagsterInstance.report_runless_asset_event is now public.
• AutoMaterializeRule.materialize_on_parent_updated now accepts an updated_parents_filter of type AutoMaterializeAssetPartitionsFilter, which allows only materializing based on updates from runs with a required set of tags.

Documentation

• Added a new guide for using Dagster Pipes with Kubernetes
• Added all OSS deployment guides to the site’s side navigation (Deployment > Open Source > Guides)
• Updated formatting in the Migrating your Dagster instance while upgrading Helm guide
• Added Experimental tags to Dagster Cloud Insights docs
• The Transitioning Data Pipelines from Development to Production and Testing against production with Dagster Cloud Branch Deployments guides have been updated to use Pythonic Resources

Dagster Cloud

• Reporting runless events and manually marking an asset as successfully materialized are no possible with “Launcher” level permissions
• [ui] Improved search and render performance of Users page, especially for large lists of users.
• [billing] Fixed issues with correctly displaying your tax ID

Bugfixes

• Fixed an error when trying to directly invoke a run status sensor when passing resources.
• [dagster-airbyte][dagster-fivetran] Fixed an issue where EnvVars used in Airbyte or Fivetran resources would show up as their processed values in the launchpad when loading assets from a live Fivetran or Airbyte instance.

Dagster Cloud

• Substantially improved performance of the Dagster insights DBT/Snowflake usage job.

New

• The OpExecutionContext and AssetExecutionContext now have a partition_keys property
• [dagster-ui] The asset graph layout algorithm has been changed to a much faster one called “tight-tree”
• [dagster-ui] The Runs table filters has a top level filter for partitions
• [dagster-dbt] dbt-core==1.7.* is now supported.

Bugfixes

• Fixed an issue where some schedules skipped a tick on the day after a fall Daylight Savings Time transition.
• Fixed a bug that caused backfill policies that execute multiple partitions in a single run not to work with dynamic partitions.
• Fixed a bug that caused an error when build_schedule_from_partitioned_job was used with a job with multi-partitioned assets and the partitions_def argument wasn’t provided to define_asset_job.
• We now raise an error early if the empty string is provided as an asset’s group name (Thanks Sierrra!)
• Fixed an issue where custom setup and teardown methods were not properly called on nested Pythonic resources.
• Added a warning message when op or asset config is passed as an argument not named config.
• [dagster-cloud] Fixed an issue where overriding the default I/O manager could break the Snowflake-dbt insights job.
• [auto-materialize] Fixed an issue where materializing an unpartitioned parent of a dynamic-partitioned asset would only result in the latest dynamic partition of that asset being requested. Now, all partitions will be requested.
• [dagster-embedded-elt] Fixed an issue in dagster-embedded-elt where sling’s updated_at parameter was set to the incorrect type
• [dagster-ui] Fixed an issue in the launchpad where selecting a partition wouldn’t correctly overwrite fields using the partition’s specific configuration

Community Contributions

• A docs fix to the testing concepts page, thanks @NicolaiLolansen!
• The schema can now be overridden per asset in DB IO managers, thanks @jrstats!

Experimental

• Improved failure recovery and retry behavior when the daemon that launches auto-materialization runs fails or crashes in the middle of a tick.
• [asset-checks] UI performance for displaying check results is improved
• [asset-checks] Removed noisy experimental warning about AssetCheckKey
• [op-concurrency] Fixed a bug where concurrency slots were not getting assigned if a run that was assigned a slot was deleted before the slot had actually been claimed during execution.
• [dagster-pipes] The PipesSubprocessClient now inherits the environment variables of the parent process in the launched subprocess.
• [dagster-pipes] Exceptions are now reported in the event log for framework components and from the external process.

Documentation

• Added a guide for using Dagster Pipes with Databricks

New

• [dagster-k8s] The PipesK8sClient will now attempt to load the appropriate kubernetes config, and exposes arguments for controlling this process.
• [ui] The launch asset backfill modal now offers a preview dialog showing the targeted asset partitions and their backfill policies when partition mapping or varying backfill policies are present.
• [asset-checks] New load_asset_checks_from_modules functions for loading asset checks in tandem with load_assets_from_modules.
• Previously, the daemon process would terminate with an error if it believed that a thread might be hanging, which sometimes caused undesirable terminations when doing large backfills or auto-materializing many assets. Now, the daemon process will emit a warning instead of terminate.
• [dagster-dbt] dagster-dbt project scaffold now uses ~/.dbt/profiles.yml if a profiles.yml is not present in the dbt project directory.
• [dagster-dbt] @dbt_assets now support PartitionMapping using DagsterDbtTranslator.get_partition_mapping.
• [dagster-dbt] Self dependencies can now be enabled for dbt models that are represented by @dbt_assets. To enable this, add the following metadata to your dbt model’s metadata in your dbt project:

meta:
  dagster:
    has_self_dependency: True

Bugfixes

• Fixed an issue where Dagster imports would throw errors when using pydantic<2.0.0 but having pydantic-core installed.
• Previously, asset backfills that targeted partitioned assets with a LastPartitionMapping dependency would raise an error. This has been fixed.
• Fixed a multipartitions partition mapping bug where a instance is not available to load partitions error was raised.
• [asset-checks] Fixed an issue with conflicting op names when using build_asset_with_blocking_check
• [ui] Viewing run logs containing very large messages no longer causes the UI to crash in Safari on macOS
• [ui] Drilling into the graph of a graph-backed asset with 4+ inputs or outputs no longer causes the asset graph to crash with a rendering error.
• [ui] On the backfill details page, clicking to a specific asset correctly shows the partitions that were materialized for that asset when partition mapping is in use.
• [ui] The Asset > Partition page now loads partition details more quickly in cases where calculating the staleness of the partition took a significant amount of time.
• Fixed a bug introduced in 1.5.0 where instances that haven’t been migrated to the latest schema hit an error upon run deletion.
• [auto-materialize] Previously, if an asset partition was skipped on a tick for one reason, and then processed and skipped on a later tick for an additional reason, only the most recent skip reason would be tracked. Now, all skip reasons are tracked.
• [dagster-dbt] Fixed an issue where if an exclude that didn’t match any dbt nodes was used in @dbt_assets, an error would be raised. The error is now properly handled.
• [dagster-dbt] When invoking DbtCliResource.cli(...) in an op, AssetMaterialization's instead of Output are now emitted.

Experimental

• Global op concurrency slots are now released in between retries of op execution failures.

Documentation

• Updated the tutorial to reflect best practices and APIs as of Dagster 1.5

Dagster Cloud

• The report_asset_observation REST endpoint for reporting runless events is now available.

New

• Dagster now supports using Pydantic 2 models for Config and Resources. Pydantic 1.10 continues to be supported.
• Added a report_asset_observation REST API endpoint for runless external asset observation events
• Dramatically improved the performance of partition-mapping, for basic hourly and daily partitions definitions
• [ui] When viewing a list of runs, you can quickly add the tag in the “Launched by” column as a filter on the list view. Hover over the tag to see the “Add to filter” button.
• [helm] The env vars DAGSTER_K8S_PG_PASSWORD_SECRET and DAGSTER_K8S_INSTANCE_CONFIG_MAP will no longer be set in all pods.
• [dagster-pyspark] build_pyspark_zip now takes an exclude parameter that can be used to customize the set of excluded files.
• [ui] Links beginning with http://, https:// in unstructured run logs (via context.log) are automatically converted to clickable links

Bugfixes

• Fixed an asset backfill bug where certain asset partitions with no targeted parents would hang indefinitely.
• Fixed a bug where the source_key_prefix argument to load_assets_from_current_module and load_assets_from_package_name was ignored
• Fixed two bugs in dagster_embedded_elt where the mode parameter was not being passed to Sling, and only one asset could be created at a time
• Fixed a bug with handing default values for Pydantic validators on Windows
• [ui] Clicking an asset with checks shows them in the asset graph sidebar, even if live data on the page is still loading.
• [ui] Reported materialization events are presented more clearly in the asset graph sidebar and in other parts of the Dagster UI.

Deprecations

• [helm] The pipelineRun configuration in the Helm chart is now deprecated. The same config can be set under dagster-user-deployments

Community Contributions

• Added setup_for_execution and teardown_after_execution calls to the inner IOManagers of the BranchingIOManager - thank you @danielgafni!
• The S3FakeResource.upload_fileobj() signature is now consistent with boto3 S3.Client.upload_fileobj() - thank you @jeanineharb!
• dbt_assets now have an optional name parameter - thank you @AlexanderVR!

Documentation

• Added a link to Dagster University to the docs landing page 🎓
• Improved readability of API docs landing page
• Removed straggling mention of Dagit from the Kubernetes OSS deployment guide

New

• Added a report_asset_check REST API endpoint for runless external asset check evaluation events. This is available in cloud as well.
• The config argument is now supported on @graph_multi_asset
• [ui] Improved performance for global search UI, especially for deployments with very large numbers of jobs or assets.
• [dagster-pipes] Add S3 context injector/reader.
• [dagster-dbt] When an exception when running a dbt command, error messages from the underlying dbt invocation are now properly surfaced to the Dagster exception.
• [dagster-dbt] The path to the dbt executable is now configurable in DbtCliResource.

Bugfixes

• Fixed a bug introduced in 1.5.3 that caused errors when launching specific Ops in a Job.
• Fixed a bug introduced in 1.5.0 that prevented the AssetExecutionContext type annotation for the context parameter in @asset_check functions.
• Fixed an issue where the Dagster scheduler would sometimes fail to retry a tick if there was an error reloading a code location in the middle of the tick.
• [dagster-dbt] Fixed an issue where explicitly passing in profiles_dir=None into DbtCliResource would cause incorrect validation.
• [dagster-dbt] Fixed an issue where partial parsing was not working when reusing existing target paths in subsequent dbt invocations.
• [ui] Fixed an issue where the job partitions UI would show “0 total partitions” if the job consisted of more than 100 assets

Community Contributions

• [dagster-duckdb] The DuckDBResource and DuckDBIOManager accept a connection_config configuration that will be passed as config to the DuckDB connection. Thanks @xjhc!

Experimental

• Added events in the run log when a step is blocked by a global op concurrency limit.
• Added a backoff for steps querying for open concurrency slots.
• Auto-materialize logic to skip materializing when (1) a backfill is in progress or (2) parent partitions are required but nonexistent are now refactored to be skip rules.
• [ui] Added 2 new asset graph layout algorithms under user settings that are significantly faster for large graphs (1000+ assets).

Documentation

• Added several pieces of documentation for Dagster Pipes, including:
  • A high-level explanation of Pipes
  • A tutorial that demonstrates how to use Pipes with a local subprocess
  • A reference for using a local subprocess with Pipes
  • A detailed explanation of Pipes, including how to customize the process
  • API references for Pipes (orchestration-side) and dagster-pipes (external process)
• Added documentation for the new experimental External Assets feature

Dagster Cloud

• Running multiple agents is no longer considered experimental.
• When the agent spins up a new code server while updating a code location, it will now wait until the new code location uploads any changes to Dagster Cloud before allowing the new server to serve requests.

New

• Alert policies can now be set on assets + asset checks (currently experimental). Check out the alerting docs for more information.
• Added a new flag --live-data-poll-rate that allows configuring how often the UI polls for new asset data when viewing the asset graph, asset catalog, or overview assets page. It defaults to 2000 ms.
• Added back the ability to materialize changed and missing assets from the global asset-graph. A dialog will open allowing you to preview and select which assets to materialize.
• Added an experimental AMP Timeline page to give more visibility into the automaterialization daemon. You can enable it under user settings
• Added a report_asset_materialization REST API endpoint for creating external asset materialization events. This is available in cloud as well.
• [dbt] The @dbt_assets decorator now accepts a backfill_policy argument, for controlling how the assets are backfilled.
• [dbt] The @dbt_assets decorator now accepts a op_tags argument, for passing tags to the op underlying the produced AssetsDefinition.
• [pipes] Added get_materialize_result & get_asset_check_result to PipesClientCompletedInvocation
• [dagster-datahub] The acryl-datahub pin in the dagster-datahub package has been removed.
• [dagster-databricks] The PipesDatabricksClient now performs stdout/stderr forwarding from the Databricks master node to Dagster.
• [dagster-dbt] The hostname of the dbt API can now be configured when executing the dagster-dbt-cloud CLI.
• [dagster-k8s] Added the ability to customize how raw k8s config tags set on an individual Dagster job are merged with raw k8s config set on the K8sRunLauncher. See the docs for more information.

Bugfixes

• Previously, the asset backfill page would display negative counts if failed partitions were manually re-executed. This has been fixed.

• Fixed an issue where the run list dialog for viewing the runs occupying global op concurrency slots did not expand to fit the content size.

• Fixed an issue where selecting a partition would clear the launchpad and typing in the launchpad would clear the partition selection

• Fixed various issues with the asset-graph displaying the wrong graph

• The IO manager’s handle_output method is no longer invoked when observing an observable source asset.

• [ui] Fixed an issue where the run config dialog could not be scrolled.

• [pipes] Fixed an issue in the PipesDockerClient with parsing logs fetched via the docker client.

• [external assets] Fixed an issue in external_assets_from_specs where providing multiple specs would error

• [external assets] Correct copy in tooltip to explain why Materialize button is disabled on an external asset.

Breaking Changes

• [pipes] A change has been made to the environment variables used to detect if the external process has been launched with pipes. Update the dagster-pipes version used in the external process.
• [pipes] The top level function is_dagster_pipes_process has been removed from the dagster-pipes package.

Community Contributions

• Override a method in the azure data lake IO manager (thanks @0xfabioo)!
• Add support of external launch types in ECS run launcher (thanks @cuttius)!

Experimental

• The Python GraphQL client is considered stable and is no longer marked as experimental.

Bugfixes

• Previously, asset backfills targeting assets with multi-run backfill policies would raise a "did not submit all run requests" error. This has been fixed.

Dagster Cloud

• The experimental dagster-insights package has receieved some API surface area updates and bugfixes.

New

• Dagster now automatically infers a dependency relationship between a time-partitioned asset and a multi-partitioned asset with a time dimension. Previously, this was only inferred when the time dimension was the same in each asset.
• The EnvVar utility will now raise an exception if it is used outside of the context of a Dagster resource or config class. The get_value() utility will retrieve the value outside of this context.
• [ui] The runs page now displays a “terminate all” button at the top, to bulk terminate in-progress runs.
• [ui] Asset Graph - Various performance improvements that make navigating large asset graphs smooth
• [ui] Asset Graph - The graph now only fetches data for assets within the viewport solving timeout issues with large asset graphs
• [ui] Asset Graph Sidebar - The sidebar now shows asset status
• [dagster-dbt] When executing dbt invocations using DbtCliResource, an explicit target_path can now be specified.
• [dagster-dbt] Asset checks can now be enabled by using DagsterDbtTranslator and DagsterDbtTranslatorSettings: see the docs for more information.
• [dagster-embedded-elt] Dagster library for embedded ELT

Bugfixes

• [ui] Fixed various issues on the asset details page where partition names would overflow outside their containers
• [ui] Backfill notification - Fixed an issue where the backfill link didn’t take the —path-prefix option into account
• [ui] Fixed an issue where the instance configuration yaml would persist rendering even after navigating away from the page.
• [ui] Fixed issues where config yaml displays could not be scrolled.
• [dagster-webserver] Fixed a performance issue that caused the UI to load slowly

Deprecations

• [dagster-dbt] Enabling asset checks using dbt project metadata has been deprecated.

Major Changes since 1.4.0 (core) / 0.20.0 (libraries)

Core

• Improved ergonomics for execution dependencies in assets  - We introduced a set of APIs to simplify working with Dagster that don't use the I/O manager system for handling data between assets. I/O manager workflows will not be affected.

  • AssetDep type allows you to specify upstream dependencies with partition mappings when using the deps parameter of @asset and AssetSpec.
  • MaterializeResult can be optionally returned from an asset to report metadata about the asset when the asset handles any storage requirements within the function body and does not use an I/O manager.
  • AssetSpec has been added as a new way to declare the assets produced by @multi_asset. When using , the does not need to return any values to be stored by the I/O manager. Instead, the should handle any storage requirements in the body of the function.

• Asset checks (experimental) - You can now define, execute, and monitor data quality checks in Dagster [docs].

  • The @asset_check decorator, as well as the check_specs argument to @asset and @multi_asset enable defining asset checks.
  • Materializing assets from the UI will default to executing their asset checks. You can also execute individual checks.
  • When viewing an asset in the asset graph or the asset details page, you can see whether its checks have passed, failed, or haven’t run successfully.

• Auto materialize customization (experimental) - AutoMaterializePolicies can now be customized [docs].

  • All policies are composed of a set of AutoMaterializeRules which determine if an asset should be materialized or skipped.
  • To modify the default behavior, rules can be added to or removed from a policy to change the conditions under which assets will be materialized.

dagster-pipes

• Dagster pipes is a new library that implements a protocol for launching compute into external execution environments and consuming streaming logs and Dagster metadata from those environments. See https://github.com/dagster-io/dagster/discussions/16319 for more details on the motivation and vision behind Pipes.
• Out-the-box integrations
  • Clients: local subprocess, Docker containers, Kubernetes, and Databricks
    • PipesSubprocessClient, PipesDocketClient, PipesK8sClient, PipesDatabricksClient
  • Transport: Unix pipes, Filesystem, s3, dbfs
  • Languages: Python
• Dagster pipes is composable with existing launching infrastructure via open_pipes_session. One can augment existing invocations rather than replacing them wholesale.

Since 1.4.17 (core) / 0.20.17 (libraries)

New

• [ui] Global Asset Graph performance improvement - the first time you load the graph it will be cached to disk and any subsequent load of the graph should load instantly.

Bugfixes

• Fixed a bug where deleted runs could retain instance-wide op concurrency slots.

Breaking Changes

• AssetExecutionContext is now a subclass of OpExecutionContext, not a type alias. The code

def my_helper_function(context: AssetExecutionContext):
    ...

@op
def my_op(context: OpExecutionContext):
    my_helper_function(context)

will cause type checking errors. To migrate, update type hints to respect the new subclassing.

• AssetExecutionContext cannot be used as the type annotation for @ops run in @jobs. To migrate, update the type hint in @op to OpExecutionContext. @ops that are used in @graph_assets may still use the AssetExecutionContext type hint.

# old
@op
def my_op(context: AssetExecutionContext):
    ...

# correct
@op
def my_op(context: OpExecutionContext):
    ...

• [ui] We have removed the option to launch an asset backfill as a single run. To achieve this behavior, add backfill_policy=BackfillPolicy.single_run() to your assets.

Community Contributions

• has_dynamic_partition implementation has been optimized. Thanks @edvardlindelof!
• [dagster-airbyte] Added an optional stream_to_asset_map argument to build_airbyte_assets to support the Airbyte prefix setting with special characters. Thanks @chollinger93!
• [dagster-k8s] Moved “labels” to a lower precedence. Thanks @jrouly!
• [dagster-k8s] Improved handling of failed jobs. Thanks @Milias!
• [dagster-databricks] Fixed an issue where DatabricksPysparkStepLauncher fails to get logs when job_run doesn’t have cluster_id at root level. Thanks @PadenZach!
• Docs type fix from @sethusabarish, thank you!

Documentation

• Our Partitions documentation has gotten a facelift! We’ve split the original page into several smaller pages, as follows:
  • Partitions - An overview of what a partition is, benefits, and how to use it
  • Partitioning assets - Details about partitioning assets
  • Partitioning ops - Details about partitioning ops
  • Testing partitions - As described

Dagster Cloud

• New dagster-insights sub-module - We have released an experimental dagster_cloud.dagster_insights module that contains utilities for capturing and submitting external metrics about data operations to Dagster Cloud via an api. Dagster Cloud Insights is a soon-to-be released feature that shows improves visibility into usage and cost metrics such as run duration and Snowflake credits in the Cloud UI.

New

• [dagster-dbt] DbtCliResource now enforces that the current installed version of dbt-core is at least version 1.4.0.
• [dagster-dbt] DbtCliResource now properly respects DBT_TARGET_PATH if it is set by the user. Artifacts from dbt invocations using DbtCliResource will now be placed in unique subdirectories of DBT_TARGET_PATH.

Bugfixes

• When executing a backfill that targets a range of time partitions in a single run, the partition_time_window attribute on OpExecutionContext and AssetExecutionContext now returns the time range, instead of raising an error.
• Fixed an issue where the asset backfill page raised a GraphQL error for backfills that targeted different partitions per-asset.
• Fixed job_name property on the result object of build_hook_context.

Experimental

• AssetSpec has been added as a new way to declare the assets produced by @multi_asset.
• AssetDep type allows you to specify upstream dependencies with partition mappings when using the deps parameter of @asset and AssetSpec.
• [dagster-ext] report_asset_check method added to ExtContext.
• [dagster-ext] ext clients now must use yield from to forward reported materializations and asset check results to Dagster. Results reported from ext that are not yielded will raise an error.

Documentation

• The Dagster UI documentation got an overhaul! We’ve updated all our screenshots and added a number of previously undocumented pages/features, including:
  • The Overview page, aka the Factory Floor
  • Job run compute logs
  • Global asset lineage
  • Overview > Resources
• The Resources documentation has been updated to include additional context about using resources, as well as when to use os.getenv() versus Dagster’s EnvVar.
• Information about custom loggers has been moved from the Loggers documentation to its own page, Custom loggers.

New

• [ui] When using the search input within Overview pages, if the viewer’s code locations have not yet fully loaded into the app, a loading spinner will now appear to indicate that search results are pending.

Bugfixes

• Fixed an asset backfill bug that caused occasionally caused duplicate runs to be kicked off in response to manual runs upstream.
• Fixed an issue where launching a run from the Launchpad that included many assets would sometimes raise an exception when trying to create the tags for the run.
• [ui] Fixed a bug where clicking to view a job from a run could lead to an empty page in situations where the viewer’s code locations had not yet loaded in the app.

Deprecations

• Deprecated ExpectationResult. This will be made irrelevant by upcoming data quality features.

Community Contributions

• Enabled chunked backfill runs to target more than one asset, thanks @ruizh22!

Experimental

• Users can now emit arbitrary asset materializations, observations, and asset check evaluations from sensors via SensorResult.

New

• The deps parameter for @asset and @multi_asset now supports directly passing @multi_asset definitions. If an @multi_asset is passed to deps, dependencies will be created on every asset produced by the @multi_asset.
• Added an optional data migration to convert storage ids to use 64-bit integers instead of 32-bit integers. This will incur some downtime, but may be required for instances that are handling a large number of events. This migration can be invoked using dagster instance migrate --bigint-migration.
• [ui] Dagster now allows you to run asset checks individually.
• [ui] The run list and run details page now show the asset checks targeted by each run.
• [ui] In the runs list, runs launched by schedules or sensors will now have tags that link directly to those schedules or sensors.
• [ui] Clicking the "N assets" tag on a run allows you to navigate to the filtered asset graph as well as view the full list of asset keys.
• [ui] Schedules, sensors, and observable source assets now appear on the resource “Uses” page.
• [dagster-dbt] The DbtCliResource now validates at definition time that its project_dir and profiles_dir arguments are directories that respectively contain a dbt_project.yml and profiles.yml.
• [dagster-databricks] You can now configure a policy_id for new clusters when using the databricks_pyspark_step_launcher (thanks @zyd14!)
• [ui] Added an experimental sidebar to the Asset lineage graph to aid in navigating large graphs. You can enable this feature under user settings.

Bugfixes

• Fixed an issue where the dagster-webserver command was not indicating which port it was using in the command-line output.
• Fixed an issue with the quickstart_gcp example wasn’t setting GCP credentials properly when setting up its IOManager.
• Fixed an issue where the process output for Dagster run and step containers would repeat each log message twice in JSON format when the process finished.
• [ui] Fixed an issue where the config editor failed to load when materializing certain assets.
• [auto-materialize] Previously, rematerializing an old partition of an asset which depended on a prior partition of itself would result in a chain of materializations to propagate that change all the way through to the most recent partition of this asset. To prevent these “slow-motion backfills”, this behavior has been updated such that these updates are no longer propagated.

Experimental

• MaterializeResult has been added as a new return type to be used in @asset / @multi_asset materialization functions
• [ui] The auto-materialize page now properly indicates that the feature is experimental and links to our documentation.

Documentation

• The Concepts category page got a small facelift, to bring it line with how the side navigation is organized.

Dagster Cloud

• Previously, when importing a dbt project in Cloud, naming the code location “dagster” would cause build failures. This is now disabled and an error is now surfaced.

New

• Added a new tooltip to asset runs to either view the asset list or lineage

Bugfixes

• [ui] Fixed an issue where re-executing a run from a particular run's page wouldn’t navigate to the newly created run

Experimental

• [dagster-ext] An initial version of the dagster-ext module along with subprocess, docker, databricks, and k8s pod integrations are now available. Read more at https://github.com/dagster-io/dagster/discussions/16319. Note that the module is temporarily being published to PyPI under dagster-ext-process, but is available in python as import dagster_ext.
• [asset checks] Added an ‘execute’ button to run checks without materializing the asset. Currently this is only supported for checks defined with @asset_check or AssetChecksDefinition.
• [asset checks] Added check_specs argument to @graph_multi_asset
• [asset checks] Fixed a bug with checks on @graph_asset that would raise an error about nonexistant checks

New

• OpExecutionContext.add_output_metadata can now be called multiple times per output.

Bugfixes

• The double evaluation of log messages in sensor logging has been fixed (thanks @janosroden !)
• Cron schedules targeting leap day (ending with 29 2 *) no longer cause exceptions in the UI or daemon.
• Previously, if multiple partitioned observable_source_assets with different partition definitions existed in the same code location, runs targeting those assets could fail to launch. This has been fixed.
• When using AutoMaterializePolicies with assets that depended on prior partitions of themselves, updating the start_date of their underlying PartitionsDefinition could result in runs being launched for partitions that no longer existed. This has been fixed.
• Fixed an issue where auto-materilization could sometimes produce duplicate runs if there was an error in the middle of an auto-materialization tick.
• [dagster-census] A recent change to the Census API broke compatibility with
  this integration. This has been fixed (thanks @ldnicolasmay!)
• [dagster-dbt] Fixed an issue where DagsterDbtTranslator did not properly invoke get_auto_materialize_policy and get_freshness_policy for load_assets_from_dbt_project.
• [ui] Fixed a number of interaction bugs with the Launchpad config editor, including issues with newlines and multiple cursors.
• [ui] Asset keys and partitions presented in the asset checks UI are sorted to avoid flickering.
• [ui] Backfill actions (terminate backfill runs, cancel backfill submission) are now available from an actions menu on the asset backfill details page.

Community Contributions

• Typo fix in run monitoring docs (thanks c0dk)!
• Grammar fixes in testing docs (thanks sonnyarora)!
• Typo fix in contribution docs (thanks tab1tha)!

Experimental

• [dagster-dbt][asset checks] Added support to model dbt tests as Dagster asset checks.
• [asset checks] Added @graph_asset support. This can be used to implement blocking checks, by raising an exception if the check fails.
• [asset checks] Fixed @multi_asset subsetting, so only checks which target assets in the subset will execute.
• [asset checks] AssetCheckSpecs will now cause an error at definition time if they target an asset other than the one they’re defined on.
• [asset checks] The status of asset checks now appears in the asset graph and asset graph sidebar.

Dagster Cloud

• [Experimental] Added support for freeing global op concurrency slots after runs have finished, using the deployment setting: run_monitoring > free_slots_after_run_end_seconds

New

• The context object now has an asset_key property to get the AssetKey of the current asset.
• Performance improvements to the auto-materialize daemon when running on large asset graphs.
• The dagster dev and dagster-daemon run commands now include a --log-level argument that allows you to customize the logger level threshold.
• [dagster-airbyte] AirbyteResource now includes a poll_interval key that allows you to configure how often it checks an Airbyte sync’s status.

Bugfixes

• Fixed an issue where the dagster scheduler would sometimes raise an error if a schedule set its cron_schedule to a list of strings and also had its default status set to AUTOMATICALLY_RUNNING.
• Fixed an issue where the auto-materialize daemon would sometimes raise a RecursionError when processing asset graphs with long upstream dependency chains.
• [ui] Fixed an issue where the Raw Compute Logs dropdown on the Run page sometimes didn’t show the current step name or properly account for retried steps.

Community Contributions

• [dagster-databricks] Fixed a regression causing DatabricksStepLauncher to fail. Thanks @zyd14!
• Fixed an issue where Dagster raised an exception when combining observable source assets with multiple partitions definitions. Thanks @aroig!
• [dagster-databricks] Added support for client authentication with OAuth. Thanks @zyd14!
• [dagster-databricks] Added support for workspace and volumes init scripts in the databricks client. Thanks @zyd14!
• Fixed a missing import in our docs. Thanks @C0DK!

Experimental

• Asset checks are now displayed in the asset graph and sidebar.

• [Breaking] Asset check severity is now set at runtime on AssetCheckResult instead of in the @asset_check definition. Now you can define one check that either errors or warns depending on your check logic. ERROR severity no longer causes the run to fail. We plan to reintroduce this functionality with a different API.

• [Breaking] @asset_check now requires the asset= argument, even if the asset is passed as an input to the decorated function. Example:

  @asset_check(asset=my_asset)
  def my_check(my_asset) -> AssetCheckResult:
      ...
  

• [Breaking] AssetCheckSpec now takes asset= instead of asset_key=, and can accept either a key or an asset definition.

• [Bugfix] Asset checks now work on assets with key_prefix set.

• [Bugfix] Execution failure asset checks are now displayed correctly on the checks tab.

Documentation

• [dagster-dbt] Added example of invoking DbtCliResource in custom asset/op to API docs.
• [dagster-dbt] Added reference to explain how a dbt manifest can be created at run time or build time.
• [dagster-dbt] Added reference to outline the steps required to deploy a Dagster and dbt project in CI/CD.
• Miscellaneous fixes to broken links and typos.

New

• Dagster code servers now wait to shut down until any calls that they are running have finished, preventing them from stopping while in the middle of executing sensor ticks or other long-running operations.
• The dagster execute job cli now accepts —-op-selection (thanks @silent-lad!)
• [ui] Option (Alt) + R now reloads all code locations (OSS only)

Bugfixes

• Adds a check to validate partition mappings when directly constructing AssetsDefinition instances.
• Assets invoked in composition functions like @graph and @job now work again, fixing a regression introduced in 1.4.5.
• Fixed an issue where a race condition with parallel runs materializing the same asset could cause a run to hang.
• Fixed an issue where including a resource in both a schedule and a job raised a “Cannot specify resource requirements” exception when the definitions were loaded.
• The ins argument to graph_asset is now respected correctly.
• Fixed an issue where the daemon process could sometimes stop with a heartbeat failure when the first sensor it ran took a long time to execute.
• Fixed an issue where dagster dev failed on startup when the DAGSTER_GRPC_PORT `environment variable was set in the environment.
• deps arguments for an asset can now be specified as an iterable instead of a sequence, allowing for sets to be passed.
• [dagster-aws] Fixed a bug where the S3PickleIOManager didn’t correctly handle missing partitions when allow_missing_partitions was set. Thanks @o-sirawat!
• [dagster-k8s] in the helm chart, the daemon securityContext setting now applies correctly to all init containers (thanks @maowerner!)

Community Contributions

• [dagster-databricks] Migrated to use new official databricks Python SDK. Thanks @judahrand!

Experimental

• New APIs for defining and executing checks on software-defined assets. These APIs are very early and subject to change. The corresponding UI has limited functionality. Docs
• Adds a new auto-materialize skip rule AutoMaterializeRule.skip_on_not_all_parents_updated that enforces that an asset can only be materialized if all parents have been materialized since the asset's last materialization.
• Exposed an auto-materialize skip rule – AutoMaterializeRule.skip_on_parent_missing –which is already part of the behavior of the default auto-materialize policy.
• Auto-materialize evaluation history will now be stored for 1 month, instead of 1 week.
• The auto-materialize asset daemon now includes more logs about what it’s doing for each asset in each tick in the Dagster Daemon process output.

Documentation

• [dagster-dbt] Added reference docs for dagster-dbt project scaffold.

Dagster Cloud

• Fixed an issue where the Docker agent would sometimes fail to load code locations with long names with a hostname connection error.