Astro release notes
Astronomer is committed to continuous delivery of both features and bug fixes to Astro. To keep your team up to date on what's new, this document will provide a regular summary of all changes released to Astro.
Latest Astro CLI Version: 1.30.0 (Release notes)
Latest Astro Runtime Version: 12.0.0 (Release notes)
October 29, 2024
Airflow mini-scheduler turned off by default
The mini-scheduler config (AIRFLOW__SCHEDULER__SCHEDULE_AFTER_TASK_EXECUTION
) is now disabled by default. Previously enabled to optimize performance at scale, the mini-scheduler sometimes led to unexpected task state changes and failures, especially when the main scheduler got out of sync. This change reduces debugging complexity, but users with very large DAGs, many downstream dependencies, and the need to run over 10,000 tasks per hour can re-enable it for potential performance benefits by setting AIRFLOW__SCHEDULER__SCHEDULE_AFTER_TASK_EXECUTION
to True
.
Bug fixes
- Fixed a bug where only users with Organization-scoped roles could create or update Deployment alerts, even if they had Deployment alert creation and update permissions enabled.
October 22, 2024
Bug fixes
- Fixed a bug affecting dbt deploys for Deployments using the Kubernetes executor.
October 15, 2024
Additional improvements
- The Scope and Scope Entity ID columns of the Alerts and Notification Channels lists are now consolidated to the Scope column, which displays the entity name and links to the Deployment or Workspace the alert is enabled for. If the scope of the Alert or Notification Channel is the entire Organization, the Scope column value is
This Organization
. This enhancement allows you to navigate directly to the Deployment or Workspace from the list.
Bug fixes
- Fixed a bug where the Default Pod Size field was showing incorrect sizes.
October 9, 2024
Updates and changes to Astro Alerting UI
- You can now define whether a notification channel is available to specific Deployments, an entire Workspace, or to an entire Organization.
- You can create and edit alerts and notification channels of any type through the Organization Alerting page, whether they are Deployment or Organization alerts. If you do not have user permissions with privileges to make a particular type of alert, you can still access the Alerting editor in the Organization menu to create, edit, and manage both alerts and notification channels.
- Previously, you could access and edit Workspace alerts from the Alerting page in the main navigation of the Astro UI. Now, you can create and manage your Organization alerts through the Alerting page in the Organization menu, or your Deployment-specific alerts in your Deployment's Alerts page.
Bug fixes
- Fixed a bug where editing Airflow variable links in the Environment Manager would delete any links to Deployments. Now, editing Airflow variables in the Environment Manager does not affect linked Deployments.
October 1, 2024
Additional improvements
- Added Updated By and Created By columns to the Workspaces list.
- Enhanced the onboarding experience so you can use one of the template Deployments without connecting to your GitHub account.
Bug fixes
- Fixed an issue that showed incorrect pagination numbers in the table footer when searching the Alerts list.
- Resolved an issue where the
globalFilter
search paramameter would not clear if you deleted the input value entirely. - Fixed an issue with the dbt deploy rollback functionality where dbt bundles weren't updated to the rollback deploy version's bundle state.
September 24, 2024
Additional improvements
- The Universal Metrics Exporter is now generally available to Team-tier level customers. See Export metrics for information about how to export infrastructure metrics about your Airflow use on Astro to your preferred third-party observability tools.
- Updated the Alerts tab in the Astro UI to improve bulk-selecting alerts, increase the number of characters you can use when naming an alert, and allow auto-generated alert names to include the alert type in the alert name.
- Improved the display of long strings to have a truncated display, which expands to show the full string when you hover over or copy the value.
Bug fixes
- Fixed an issue where you couldn't submit Log Summary feedback for different task runs that had the same task ID.
September 17, 2024
Additional improvements
- Updated the Organization Dashboards' Cost Breakdown tab, which now displays top-level Network and Cluster costs.
Bug fixes
- Fixed an issue which prevented KPO task Pods from terminating correctly.
September 9, 2024
Scale Deployments Reliably with XL Deployment size and standalone DAG processor
As your team and Airflow use cases grow, scaling up your Airflow environment reliably can be a challenge. Complex, dynamically-generated DAGs, sub-optimal DAG parsing practices, or a growing business that requires a larger data pipeline can strain DAG processing and threaten your Airflow scheduler’s availability.
Astro Hosted Deployments now support high-scale environments more reliably by separating the DAG processor from the scheduler. You can take advantage of this Airflow best practice in Astro Hosted Deployments with Medium and larger sizes. By separating the DAG processor from the scheduler, Astro protects your core scheduling function from competing with the DAG processor for resources. This separation ensures scheduler availability for your most complex use cases and improves overall DAG parsing performance.
In addition to this change, you can now create Extra Large size Deployments to confidently run your largest workloads on Astro. The Extra Large Deployment provides 8 vCPU and 16 GiB of memory and comes with two DAG processors to support your largest and most complex workloads, ranging from dynamically-generated DAGs to high-scale pipelines that can grow alongside your business.
This update is available on Runtime version 9.7.0 and greater; if you are running a lower Runtime version, upgrade your Runtime version. To learn more about this feature, read more about Deployment scheduler resources.
Astro UI improvements
Improved the navigation and design for managing clusters, editing settings, and accessing analytics for Organizations and Deployments.
These changes include:
- A new section in the main menu to include an Organization section with links to important resources for administering your Astro Organization. These include links to your Organization Settings, Organization Dashboards, and quick access to your cluster settings.
- A dedicated Deployment Analytics page in your detailed Deployment information, which now includes improvements to the tooltips and metrics visualization design. See Deployment metrics for more information about the available metrics.
- Your Deploy History now appears on the Overview page of your Deployment. See Deploy history for more information about how to view your code deploy history.
Create and manage Airflow variables with the Environment Manager
You can now create and manage Airflow variables for Deployments across your Workspace through the Astro Environment Manager. This allows you to quickly and securely create Airflow variables once and share them to multiple Deployments without having to set up your own secrets backend.
See Create Airflow variables in the Astro UI for more information.
September 3, 2024
Additional improvements
- Made updates to the Astro UI for editing Astro Alerts, to clarify whether you are discarding your edits to an alert, or discarding the alert.
Bug fixes
- Fixed an issue that blocked some users from pushing project images to the Docker registry.
August 27, 2024
Additional improvements
- You can now give Airflow Dataset Create and Delete permissions to custom roles. This capability is provided to Deployment Admin, Workspace Author, Workspace Operator, and Workspace Owner default roles. Previously only Workspace Owner and above had access to this permission set. See more information about User Permissions and Custom Deployment Roles.
- If your Astro Deployments run on AWS or GCP clusters, you can now see the Account ID of the clusters in the Cluster Details page in your Organization Settings.
- For Deployment alert emails, the sender now shows as Astro. It previously showed the sender as Postmaster.
- You can now see Deployment details while Astro creates your Deployment, where only the specific actions that are not available until after creation is complete are disabled.
August 20, 2024
Additional improvements
- Added the ability to specify Custom Workload Identity at Deployment creation time when creating an AWS Deployment.
- Improved the descriptions of default hibernation schedules in the UI.
- Added more detail in descriptions for code deploy failures in the Deploy History page.
Bug fixes
- Fixed an issue where you could not use global text search for DAGs in your Alerts.
- Resolved a problem where search query parameters might be deleted when using pagination.
August 13, 2024
Additional improvements
- Improved the descriptions of Deployment Health Alerts to provide more specific information about performance thresholds.
Bug fixes
- Fixed an issue where DAG lists didn't update when you switched Deployments in the Astro UI.
- Fixed a bug where the cursor would disappear while typing in certain fields in the Astro UI.
August 7, 2024
Customize labels on your metrics export
You can now add Key:Value pair labels to metrics that you export using the Universal Metrics Exporter. This allows you to tag and record metrics coming from specific Deployments or Workspaces. See Export metrics from Astro for more information.
Additional improvements
- Added a new template demonstrating how to run dbt projects on Astro using Cosmos when you're onboarding. See Get started on Astro to set up a trial account and try out a template.
Bug fixes
- Fixed an issue where a user could not bulk select alerts if they had different names but the same configurations.
July 30, 2024
New alerts for Deployment health incidents
This release introduces four new alert types that correspond to four of the existing Deployment health incident types. These new Deployment Health Alerts allow Astro to proactively notify you when Deployment health issues arise.
Using Deployment health alerts, you can:
- Improve alerting coverage beyond DAG and task failures to address infrastructure-level incidents that are otherwise difficult to monitor.
- Proactively monitor Deployment health and take immediate remediation actions through email, Slack, and PagerDuty to reduce mean time to resolution.
- Share Deployment health visibility across teams.
See Astro Alerts: Trigger Types and Deployment Health Incidents for more information.
Additional improvements
- Now, when you link directly to your Astro or Airflow UI, any link previews successfully have improved visuals and include metadata information.
- Improved the Astro Alerts UI to streamline creating notification channels and alerts across DAGs and Deployments. See Astro Alerts for setup information.
- Added an example Universal Metrics Export dashboard configuration file for Grafana Cloud. See Export metrics from Astro for setup instructions and Grafana example for a configuration example.
- Organization Dashboards are now generally available to Enterprise-tier customers. See View Organization Dashboards for more information.
Bug fixes
- Fixed a bug where the Astro UI would show an error in your Deploy History instead of your Runtime Version, if the Runtime version is
yanked
. - Fixed an issue when creating an alert where you couldn't select tasks with the same
task_id
across multiple selected DAGs.
July 24, 2024
Run your first DAG on Astro with GitHub Integration
Now, when you first set up your Astro account, you can choose to connect your GitHub account and run a sample Astro project template and DAG. Previously, to run your first DAG on Astro, you needed to either download and use the Astro CLI or work with GitHub Actions. Now, the onboarding process allows you to customize your first experience with Astro to focus on your use case, whether that's Business Operations, Generative AI, or learning Airflow, and set up a GitHub Integration to quickly clone and deploy the example DAGs. Try it now with Start Astro.
Work with dbt projects on Astro
You can now deploy dbt code to Astro quickly and independently of your DAG or image deploys. Data build tool (dbt) core is an open source tool for data transformation that uses SQL to transform your data instead of Python in your DAGs. You can also use both Cosmos, an open source tool that integrates dbt models into Airflow and treats them like DAGs, and dbt deploys on Astro to have unparalleled visibility into your dbt tasks with the Airflow UI and a streamlined code deploy process.
See Deploy dbt projects for more information about how to get started.
Additional improvements
- Improved how the GitHub Integration in the Astro UI links directly to the Astro Project in a GitHub repository.
- (Astro Hosted only) The customer managed workload identity setting for Deployments is now generally available. This allows you to grant Astro Deployments all of the permissions of an AWS IAM role. See Attach an IAM role to your Deployment for detailed information.
- The ability to deploy automatically from GitHub using the official Astro GitHub integration is now generally available. See Deploy code with GitHub for setup steps.
- Astro Runtime version 11 is now classified as Long Term Support (LTS). This means that instead of being supported until October 2024, Astro will support Runtime version 11 until October 2025. See the Astro Runtime maintenance and lifecycle schedule for more information, including information about restricted versions.
Bug fixes
- Fixed some bugs in the Astro UI for defining ephemeral storage to show storage limits, fix maximum pod size for storage, and add storage for default pod size.
July 16, 2024
Additional improvements
- Added the following metrics to Metrics Export:
- The new
airflow_executor_open_slots
,airflow_dagrun_dependency-check
, andairflow_dagrun_dependency-check.<dag_id>
metrics allow you to collect metrics about your DAGs and executor status. - Monitor task execution with
kube_pod_container_resource_limits
. This new metric enables you to track resource use against the configured limits so you can understand if task execution meets your configured CPU, memory, or storage limits for your Celery Workers or Kubernetes Executor and KubernetesPodOperator pods.
- The new
- Added new suporting documentation for the Astro Terraform Provider, including a Getting Started guide and code examples for common uses. Read Astro Terraform Provider for more information.
Bug fixes
- Fix an error where start times in the UI were different from actual DAG run trigger times.
July 9, 2024
Additional improvements
- Fixed a bug that prevented adding a date input for a DAG filter in the Astro UI.
July 2, 2024
Export metrics about your Astro Deployments to observability tools
You can now export comprehensive, operational metrics about the performance of your Astro Deployments to third-party observability tools, such as New Relic, using the new Universal Metrics Exporter. This new feature allows you to configure a Prometheus endpoint to export metrics using the Prometheus data model, which means you can integrate your Astro observability metrics directly into your existing monitoring tools. See Export metrics from Astro for setup instructions.
Added ephemeral storage metrics to Astro Deployment Analytics
To assist you in determining the amount of custom ephemeral storage to configure for your workers and schedulers, you can now use Deployment Analytics to see relevant usage metrics. These include:
- Ephemeral Storage Usage metric that shows a % of usage against the configured limit for your Celery Workers, KubernetesPodOperator, Kubernetes Executor, and Scheduler.
- Dynamic Y-axis Scaling to the Celery Worker, KPO/KE, and Scheduler Deployment analytics panel to include dynamic zooming.
See Deployment Analytics for more information.
Additional improvements
- The
Scheduler heartbeat not found
Deployment health incident is downgraded toWarning
fromCritical
.
June 25, 2024
Customer managed workload identity for AWS
The Customer Managed Identity Deployment setting is now available on AWS. This means that you can now assign an existing workload identity and AWS IAM role to your Airflow Deployments on Astro. When you use this setting, your Deployment uses the identity to assume the permissions of your IAM role and gain secure access to your data services. With this feature, you can:
- Re-use or share a customer managed identity across many Deployments, either ephemeral or static.
- Leverage existing identities when migrating from MWAA or open source Airflow environments
This can reduce friction when migrating to Astro. See Attach an IAM role to your Deployment for detailed information.
Self-healing workers automatically address stuck queued tasks
A new improvement to the Astro Data Plane now automatically identifies when Celery workers are online and healthy, but not actually processing new tasks. When this happens, it looks like many tasks are stuck in a queued
state. With this new feature, you will experience a lower frequency of tasks stuck in a queued
state, which causes performance and reliability issues for your Airflow implementation.
Self-healing workers use the following process:
- It first identifies workers that both have tasks stuck in a
queued
state for an extensive time period and are in Deployments where the concurrency available means that tasks should not be queued. - The healer then kills catatonic workers that are not running tasks or shifts workers that are still running tasks into a warm shutdown period.
- After the catatonic worker is shut down, a new healthy, worker comes online and resumes tasks.
This feature is automatically enabled for the Astro Hosted infrastructure and does not require any action.
Additional Improvements
- Improved the formatting for how IP addresses are listed in the Astro UI to make it easier to copy and paste them.
Bug fixes
europe-west6
is no longer available as a region for dedicated clusters on GCP.
June 18, 2024
Additional Improvements
- Moved the Environment variable tab in Deployment Settings to the Environment tab, and renamed it to Environment Variables. This helps disambiguate between Environment Variables and Airflow Variables in the Astro UI. Refer to Manage environment variables for more information about the different methods to set up and manage Environment variables.
Bug fixes
- Fixed a bug where a duplicate Customer managed identity option displayed when configuring Deployment settings.
June 11, 2024
Restrict Astro access to specific IP Address ranges with IP Access list
Organization owners can now limit the IP addresses of users to control who can access the Astro UI or programmatically work with the Astro API for their organization. This allows you to restrict access to your Astro Organization to users on a VPN or to specific network ranges. See Set up IP Access list.
Additional improvements
- When using the GitHub integration, you can now retry a failed deploy or trigger a Git deploy directly from your Deployment settings page in the Astro UI. See GitHub Integration for more information.
- Deployments at the end of an Astro trial now automatically hibernate and are not immediately deleted. If your trial has ended, add a credit card number to your Organization to access your Workspaces and wake your hibernating Deployment. While in hibernation, Deployment settings and configurations are preserved. If you do not enter a payment method within 30 days, your hibernating Deployments and all corresponding data are deleted. See Start a trial for more information.
- The ability to create, update, and delete Deployment API tokens is restricted to users with the Workspace Owner role. Read Workspace user permissions for more information.
Bug fixes
- Fixed an issue where the Organization Settings Dashboard failed to load correctly.
June 6, 2024
API keys are no longer supported
As of June 1, 2024, Deployment API keys are no longer supported. Replace your API keys with Deployment API tokens, confirm successful operation with your API tokens, and then delete your API keys.
Configure ephemeral storage on worker Pods
You can now customize the amount of ephemeral storage for data intensive workloads on Celery, Kubernetes, and KubernetesPodOperator workers. Previously, to accommodate larger workloads, you needed to integrate an external object storage or database tool to process large datasets within a single task. Now, you can customize the ephemeral storage when creating or updating your Deployment so that all data processing happens directly in your task Pod.
You are only charged for requested resources which are greater than the minimum defaults for each worker type:
- Celery worker: 10 GiB minimum by default. 100 GiB maximum.
- Kubernetes executor/ Kubernetes pod operator: 0.25 GiB minimum by default. 100 GiB maximum.
Automate Airflow, resource, and infrastucture management with the Astro Terraform Provider
You can now use Terraform to automate managing resources and changes to large organizations and Airflow infrastructure with the Astronomer Terraform provider package. The provider is available through both the Terraform Registry and a public Github repository, where you can review the provider's code, make issues, and create pull requests.
Refer to the Astronomer Terraform Provider docs in the Terraform registry or the GitHub repository for more information.
Additional improvements
- The Astro UI now validates cron expressions for development Deployment hibernation schedules and shows commonly used hibernation schedules that you can enable or disable. Refer to Create a hibernation schedule for more information.
- If you don't already have a GitHub repository, you can now create one when you authorize the GitHub Integration from the Astro UI. Previously, you could only connect Astro to existing repositories.
Bug fixes
- Fixed an issue where you could not use Airflow connection testing with Azure Managed Identities on Astro.
May 30, 2024
Additional improvements
- Dedicated clusters are now available only to Team tier customers and above.
- Added a centralized docs reference page that lists all open source Apache Airflow provider packages and their versions for each Astro Runtime version. See Provider package reference docs.
Bug fixes
- Fixed an issue where you could make changes to clusters and Deployments on an inactive Organization.
- The default storage resources for the Kubernetes executor and KubernetesPodOperator Pods are now enforced with
0.25Gi
instead of10Gi
in some cases. You can still customize the resource allocation for Kubernetes pods depending on your needs. See Configure Kubernetes Pod Resources
May 22, 2024
The Astro GitHub integration is now in Public Preview
The ability to deploy automatically from GitHub using the official Astro GitHub integration is now in Public Preview.
The Astro GitHub integration is a new way to automatically deploy code from a GitHub repository to Astro by merging pull requests or making commits directly to specific branches, without needing to configure a GitHub Action. Additionally, the GitHub integration displays Git metadata directly in the Astro UI, including Git commit descriptions and gives you greater visibility into the status and logs of individual code deploys.
See Deploy code with the Astro GitHub integration for more information.
Bug fixes
- Fixed an issue where users with custom roles could see roles that could not be assigned in the Astro UI.
May 15, 2024
Additional improvements
- Airflow connections that you configure through the Astro UI environment manager are now mounted to Deployment schedulers, meaning that scheduler processes can now make use of these Airflow connections.
Bug fixes
- Fixed an issue where you couldn't update users who were added through SCIM but didn't belong to an Organization.
- Fixed an issue where a user with a custom role could create API Tokens, users, or teams with greater permissions than their own.
May 8, 2024
Updates to address ranges for dedicated clusters on Google Cloud Provider
Astro on GCP Dedicated Clusters uses source network address translation (SNAT) that performs many-to-one IP address translations for connections to your data sources and defaults secondary ranges to RFC 6598 address space (non-standard Private IP addresses), to minimize the risk and concern with IP overlap and exhaustion. Your target data sources will see connections from Astro using the VPC Subnet Range when using private networking, like VPC Peering or VPN. If you want to configure private connectivity, ensure the default subnet and peering ranges do not overlap with your target data source network when you're creating your dedicated cluster. See Create a dedicated Astro cluster for more details.
Improvements to Astro performance
As part of continued investment in the reliability, performance, and scalability of Astro, Astronomer is embarking on a migration of public and private image registries. Astro Runtime clusters will benefit from more performant, globally distributed and geo-replicated image registries, with built-in registry resilience if a regional outage occurs. This is in addition to the previous release of the registry cache local to every Astro cluster.
No end user or task runtime change or impact is expected as part of the backend cutover during the week of May 6th, 2024.
Additional improvements
- You can no longer create Deployments using Astro Runtime versions marked as
yanked
inhttps://updates.astronomer.io/astronomer-runtime
, even if your Organization has enabled creating Deployments with deprecated Runtime versions. These versions of the Astro Runtime have known issues and should not be used. For more information, see Restricted Runtime Versions.
Bug fixes
- Fixed a bug where multiple users could not access Organization Dashboards simultaneously.
April 30, 2024
Deploy automatically from GitHub using the official Astro GitHub integration
The Astro GitHub integration is a new way to automatically deploy code from a GitHub repository to Astro without needing to configure a GitHub Action. Compared to using GitHub Actions, the Astro GitHub integration:
- Allows you to enforce software development best practices without maintaining custom CI/CD scripts.
- Enables developers to iterate on DAG code quickly.
- Shows Git metadata directly in the Astro UI, including Git commit descriptions.
- Gives you greater visibility into the status and detailed logs of an individual deploy.
See Deploy code with the Astro GitHub integration for more information.
April 23, 2024
Restrict a custom Deployment role to specific Workspaces
You can now restrict the use of a custom Deployment role to specific Workspaces. Use Workspace role restriction when some Workspaces in your Organization have different requirements for how users interact with Deployments. See Restrict a custom Deployment role to specific Workspaces for setup steps.
Additional improvements
-
The custom Deployment roles feature is now generally available.
-
You can now promote a development Deployment to a production Deployment by switching off the Development Mode toggle in the Deployment's configuration.
-
Workspace Members can now see and use custom Airflow menu items. To give a custom role this permission, you can add
deployment.airflow.customMenu.get
to the role's permissions list. This permission works only on Deployments running Astro Runtime 9 or later.Note that you might have to modify the code for your menu item plugins to make them work on Astro. See Appbuilder menu items for more information.
-
You can now filter the Workspaces and clusters lists in the Astro UI by name.
Bug fixes
- To improve the reliability of data lineage for customers who leverage it, data lineage is now a private preview feature that can be enabled upon request. To reenable data lineage for an Astro Organization, reach out to your account team.
April 16, 2024
Bug fixes
- Fixed an issue where you couldn't grant a custom Deployment role to a Deployment API token using the Astro API.
April 9, 2024
Additional improvements
- You can now manage custom Deployment roles using the Astro API.
- Improved the time it takes to load Deployment analytics in the Astro UI.
- You can now view info-level incidents from the Deployment health status indicator in the Astro UI.
- It is now possible for workers to use up to 6400 CPUs and 12800 GiB of memory on a single Deployment.
Bug fixes
- Fixed an issue where you couldn't configure boolean values for Airflow connections in the Astro UI.
- Fixed an issue where Airflow connections configured through the Astro UI did not work with deferrable tasks.
April 3, 2024
Additional improvements
- Deployment health incidents are now generally available.
- Deployment hibernation is now in public preview.
Bug fixes
- Fixed an issue where connections configured in the Astro UI were not mounted to triggerer Pods, resulting in failed runs for tasks that use deferrable operators.
- Fixed an issue where a field in the Snowflake - Private Key (Content) connection type was not parsed correctly.
- Fixed an issue where Organization Members could delete and update Astro alerts.
- Fixed an issue where the Astro UI didn't show the correct amount of available ephemeral storage for the default worker queue.
- Removed additional dependencies to make Astro more resilient to Quay outages.
March 26, 2024
Refactored Astro API documentation
Astro API documentation is now hosted at https://www.astronomer.io/docs/api. In the new API documentation center, you can:
- Format and test API requests directly in your browser.
- Export requests to Python, Javascript, and curl.
- View weekly changelogs for the API.
Create Deployments with deprecated versions of Astro Runtime
You can now use the Astro API to create Deployments with deprecated versions of Astro Runtime. Using deprecated Astro Runtime versions is sometimes necessary if you're migrating existing Airflow environments to Astro, or if you need to maintain deprecated environments for testing purposes.
Note that this feature is disabled by default. To use this feature, reach out to your account team and request for the feature to be enabled. See Run a deprecated Astro Runtime version for more information.
New GCP database instance types available
You can now use the following node instance types for database instances in GCP clusters:
- XLarge Compute Optimized (24 CPU, 48 GiB MEM)
- XXLarge Compute Optimized (32 CPU, 64 GiB MEM)
See GCP Hybrid cluster settings for a list of all available database instance types.
Additional improvements
- The Astro UI now includes Learning Bytes for features that are not yet configured within your Organization.
- The Astro UI now loads the status for DAG and task runs more quickly.
Bug fixes
- When you change a worker type for an existing worker queue, the Astro UI no longer resets the worker queue's concurrency configurations.
- Fixed an issue where the Astro API did not return the correct value for
IsHibernating
when you queried Deployment information.
March 19, 2024
New Azure regions available on Astro Hosted
You can now create Hosted dedicated clusters in the following Azure regions:
centralus
westus3
southcentralus
See Astro Hosted resource reference for more information.
Custom Deployment roles are now in Public Preview
The ability to customize Deployment-level permissions with Deployment roles is now in Public Preview.
You can additionally use the new Workplace Accessor and Deployment Admin roles to define which users have access to specific Deployments in your Workspace. See more in User permissions reference and and Create and assign custom Deployment roles.
Export data from reporting dashboards using webhooks
In addition to exporting report data with downloads or email, you can now export reporting data using webhooks. Use webhooks to send your reporting data to services such as Segment, Airtable, or Marketo.
Note that webhook exports are a Sigma feature in beta and might experience behavior changes.
Additional improvements
- You can now use the new Credits tab in the Organization Billing page to see your credit balance.
- On Astro Hybrid, the maximum worker concurrency on a worker queue has increased from
64
to256
.
March 7, 2024
Reporting dashboards are now in public preview
Organization dashboards are now in Public Preview to use for examining key metrics across your Organization.
You can also export data from dashboards in the format of your choice. Exports can be triggered on a regular schedule or as an alert when specific criteria are met in your data. Export reporting data to share with other team members or to keep a record of key performance indicators. See Export reporting data for more information.
Customize Deployment-level permissions using Deployment roles
Custom Deployment roles are a new way to define granular permissions for Astro users. For the first time, you can set a user's permissions at the Deployment level and define which specific parts of a Deployment they can access or modify. Use custom Deployment roles to have users collaborate in the same Workspace with only the minimum permissions they require. See Customize Deployment roles for more information.
New Deployment registry cache to improve resiliency
Deployments now include a cache of Astronomer's image registry that stores the current Astro Runtime image for your Deployment. Because Deployments now always have access to their running image, image registry outages should no longer result in failed DAG runs.
Additional improvements
- Removed nonfunctional network usage per Pod metrics from the Deployment Analytics page.
- The end-of-life date for Deployment API keys is June 1, 2024.
- The Cloud UI has been renamed to the Astro UI across all help text and documentation.
- Due to a minor change to Astronomer cluster architecture, you can no longer add custom tags to Hybrid clusters on AWS.
Bug fixes
- Fixed an issue where CPU usage per Pod metrics did not render correctly in the Deployment Analytics page.
- Fixed an issue where the Astro UI didn't show all available Teams when selecting Teams to add to a Workspace.
February 27, 2024
Use a custom service account to authorize Deployments to GCP
You can now attach a custom GCP service account to your Deployment to grant the Deployment all of the service account's permissions to your cloud. Using a custom service account provides the greatest amount of flexibility for authorizing Deployments to your cloud. For example, you can use existing service accounts on new Deployments, or your can attach a single service account to multiple Deployments that should all have the same level of access to your cloud. For setup steps, see Authorize Deployments to your cloud.
Bug fixes
- Fixed an issue where the Astro API failed to list Deployments after you deleted a hibernation override setting.
February 21, 2024
New worker types
Astro Hosted Deployments now support A120 and A160 workers, which include enough CPU and memory to handle the most resource-intensive tasks in your DAGs. See Astro Hosted resource reference for more information about each worker type.
New platform variables to improve Celery executor reliability
Astro Deployments now have the following environment variables set by default. This is to ensure that the Celery executor doesn't freeze when it attempts to connect with a Pod in the Celery backend that has unexpectedly terminated.
AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_TIMEOUT=30
AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_CONNECT_TIMEOUT=5
AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__SOCKET_KEEPALIVE=True
AIRFLOW__CELERY_BROKER_TRANSPORT_OPTIONS__RETRY_ON_TIMEOUT=True
For more information about each of these variables, see Platform variables
Ephemeral storage limit on schedulers
Astro now limits that amount of ephemeral storage in a scheduler to 5Gi. If a scheduler attempts to use more than 5Gi of ephemeral storage, it will be terminated.
It is rare for schedulers to require more than 5Gi of ephemeral storage. If your schedulers start to terminate after this update, ensure the following:
- Your Deployment is not producing large amounts of temporary files without cleaning them up.
- If you're using the BingAds Python SDK, your Deployment uses version 13.0.14 or later. Earlier releases include a bug that generates large amounts of temporary files.
Bug fixes
- You can no longer create Teams using the Astro API in an Organization that has SCIM provisioning enabled.
- Astro API responses have been standardized to always return cloud provider names in uppercase (for example,
AWS
).
February 13, 2024
New Astro reporting dashboards show metrics for Deployments across your Organization
This feature is in Private Preview. Please reach out to your customer success manager to enable this feature.
The new Dashboards page includes a suite of dashboards that you can use to asses the performance of Deployments and DAGs across your entire Organization. Each dashboard focuses on a different aspect of your data pipelines to show you opportunities for cost and performance improvements. You can additionally configure Astro to send you alerts when a given metric reaches a specific threshold. See Organization Dashboards for summaries of each available dashboard.
Additional improvements
- When you submit a support request from the Astro UI, you must now define an Active Engagement Period when you or a member of your team can engage with a member of Astronomer support.
- Workspace Members can now access the Clusters view in the Airflow UI for a Deployment.
Bug fixes
- Fixed an issue where network connections between clusters could be disrupted occasionally.
- When you retrieve information about a Deployment through the Astro API, the API now returns an empty value for
EnvironmentVariables
if the Deployment has no environment variables. - Deleting a Workspace through the Astro API now deletes all Astro Cloud IDE projects associated with the Workspace.
- Fixed an issue where you could not clear optional fields in a Deployment's configuration using the Astro API.
February 6, 2024
Bug fixes
- Fixed an issue where all Deployment task logs included the error
Not exporting configs to configmap...
January 30, 2024
New messages for Deployment health status
This feature is in Public Preview.
Astro now automatically monitors Deployments and notifies you when a Deployment isn't running as expected, such as when it can't detect a heartbeat in a scheduler. These notifications, known as Deployment incidents, appear in your Deployment's health status in the Astro UI.
See Deployment health incidents to learn more about each available incident type and how to address them.
Additional improvements
-
You can now access Ask Astro from the Astro UI Help menu:
-
User role titles are now consistently formatted across the Astro UI.
January 25, 2024
Self-service VPC peering and route management for AWS
This feature is in Public Preview.
You can now configure a network connection between Astro and an AWS VPC without contacting Astronomer support. Astro automatically handles creating a connection request and provides instructions for completing the setup yourself. After you create a VPC connection, you can configure routes whenever you need to connect to an additional service in your external VPC. See Create a private connection between Astro and AWS.
Additional improvements
- There is no longer a minimum on the amount of CPU and memory that you can request for Kubernetes Pods.
Bug fixes
- Fixed an issue in Astro Hybrid where refreshing the browser could occasionally reset a worker queue's worker type setting.
January 16, 2024
Additional improvements
- The Astro UI Deployment analytics page now shows CPU Usage Per Pod (%) and Memory Usage Per Pod (MB) as a percentage of your total available resources rather than the resources of a single worker Pod, such that these metrics will never show Deployment resources usage as exceeding 100%.
- The maximum value for worker queue Max # of workers has increased from 30 to 100.
January 9, 2024
Additional improvements
- When you create a Deployment, the Astro UI now shows the Airflow version of your Deployment instead of the equivalent Astro Runtime version.
- Enabled point-in-time restore (PITR) on Hybrid GCP clusters to improve resiliency to outages. Note that this might result in increased costs for cloud storage.
Bug fixes
- Fixed an issue where you occasionally couldn't access the Airflow UI for a Deployment with the error "No healthy upstream".
- Fixed an issue where deleting a user with no Workspace membership from an Organization would affect how their Workspace membership appeared in other, unrelated Organizations.
- Fixed an issue where the Open in Airflow button on the DAG details page in the Astro UI did not open the Airflow UI as expected.
December 20, 2023
Bug fixes
- Fixed an issue where creating an alert for a DAG through the Astro API would apply the alert to the incorrect Deployment.
- Fixed an issue where Deployment worker Pods could crash when running the Kubernetes executor.
- Fixed an issue where Deployment API key expiration dates were not applied correctly if you configured multiple API keys at once.
- Fixed an issue where logging features could be disrupted if you set
AZURE_CLIENT_ID
as an environment variable. Note that this fix applies only to Astro Runtime 10 and later.
December 12, 2023
Bug fixes
- Fixed an issue where the Astro UI would produce an error if you updated an environment variable on an Astro Hybrid Deployment running the Kubernetes Executor.
December 6, 2023
Additional improvements
- The Astro Environment Manager is now generally available. This feature allows you to create and manage Airflow connections in the Astro UI.
Bug fixes
- Fixed an issue where DAG code that appeared in the Airflow UI did not roll back when you rolled back a Deployment, even though the running code was successfully rolled back.
- Fixed an issue where you could not view billing information from the Astro UI when you installed Astro through the Azure Marketplace.
- Fixed an issue where the Astro UI would produce a console error when a user accessed their Workspace list.
November 30, 2023
Support for Microsoft Entra Workload ID
You can now use Microsoft Entra Workload ID to authorize Deployments to resources in Azure. Workload identity is a simple and secure way to authorize access external resources, as it doesn't require creating or storing long-term credentials. To set up Microsoft Entra Workload ID, see Authorize Deployments to cloud resources.
Bug fixes
- Removed the ability to create Hybrid Azure clusters in
eastasia
because some workload identity features aren't supported in this region.
November 16, 2023
Install Astro from the Azure Marketplace
Astro is now available as an Azure Native ISV Service. If your team is considering Astro and you use Azure, Astronomer recommends installing Astro from the Azure Marketplace because:
- You can manage billing from the Azure Portal.
- Microsoft Entra ID is pre-configured for all Organizations.
- It's easier to create Astro resources and get started directly from Azure.
See Install Astro from the Astro marketplace for setup steps. To learn more about Astronomer's partnership with Microsoft, see Introducing Apache Airflow™ on Astro – an Azure Native ISV Service.
Create Airflow connections in the Astro UI and link them to Deployments
You can now create Airflow connections in the Astro UI through the new Environment Manager menu. The Environment Manager lets you create Airflow connections directly in the Astro UI and stores all connections in an Astro-managed secrets backend. You can then share connections between Deployments and set default connections so that your team members always have access to external resources when they create new Deployments. See Create Airflow connections in the Astro UI.
Note that this feature is currently available only for Deployments running the Celery executor.
Trigger a DAG from an Astro alert
You can now configure Astro alerts to trigger any DAG in your Workspace through the Airflow REST API. You can configure the triggered DAG to complete any action, such as sending an alert through a custom communication channel or writing data about the incident to a table.
New Azure regions available on Astro Hosted
You can now create Hosted dedicated clusters in the following Azure regions:
eastus
canadacentral
uksouth
brazilsouth
centralindia
francecentral
japaneast
See Astro Hosted resource reference for more information.
November 7, 2023
Roll back Deployments to previous versions of your code
This feature is in Public Preview.
Astro now maintains snapshots of your past deploys, including your Deployment image and DAG code, for the previous three months. If you need to quickly revert a Deployment back to a working version of your code, you can roll back to a past deploy from the Deploy History page in the Astro UI.
Deploy rollbacks are a powerful safety mechanism to ensure that your production pipelines continue to run when something unexpected happens after a deploy. See Roll back to a past deploy for more information and configuration steps.
Bug fixes
- Fixed an issue where, you could inadvertently open the support request window if you opened a DAG that included "Support" in its name from the DAGs view. As a result of this change, the support request window URL has been updated from
https://cloud.astronomer.io/support
tohttps://cloud.astronomer.io/open-support-request
. - Fixed an issue where the Deployment configuration menu in the Astro UI didn't always show your Deployment's current configuration.
- Fixed an issue where you could not create a Deployment with some stable Runtime versions using the Astro Platform API.
October 31, 2023
Deployment API keys are now deprecated
Deployment API keys have been officially deprecated in favor of Deployment API tokens. This means:
- You can't create new Deployment API keys.
- If a Deployment has no configured Deployment API keys, the API keys tab will not appear.
- You can continue using existing Deployment API keys until a future end-of-support date.
Additional improvements
- Workspace Operators can now create, update, and delete Deployment API tokens.
Bug fixes
- Fixed an issue where a Deployment's Updated By field was not updated if you transferred the Deployment.
October 24, 2023
New Azure regions available on Astro Hosted
You can now create Deployments in standard clusters in the following Azure regions:
eastus2
westus2
westeurope
See Astro Hosted resource reference for more information.
Bug fixes
- Fixed an issue where Deployment API tokens weren't deleted after their associated Deployment was deleted.
- Fixed an issue where you could not create a Deployment with some available Runtime versions using the Astro Platform API.
October 17, 2023
Additional improvements
- You can now view deploy history for both Hosted and Hybrid Astro Deployments in the Astro UI. For more information, see View deploy history.
Bug fixes
- Modified behavior for Astro Hosted so that KubernetesExecutor and KubernetesPodOperator pods running in-cluster have equivalent resource requests and limits. If you do not configure them to have equivalent resource requests and limits, Astro modifies them to the become the limits. Previously, the DAG deploy would fail if
resources
did not equallimits
.
October 10, 2023
Deployment API tokens
Deployment API tokens are now generally available and replace Deployment API keys as the most secure and customizable way to programmatically update Deployments. This includes using them to deploy code and update environment variables.
See Deployment API tokens to learn how to create and manage Deployment API tokens.
Deployment API tokens are a direct replacement for Deployment API keys, which are now supported only on a limited basis on Astro.
After October 31, 2023, you will not be able to create new API keys. While you can still continue to use and manage existing Deployment API keys, Astronomer will soon require using Deployment API tokens.
New Edit Deployments in the Astro UI
Editing Deployments in the Astro UI has a new, consolidated flow. All of the configuration options are now editable in a single form, similar to Deployment creation, instead of spread across multiple forms. See Deployment Settings for a detailed description of how to create, update, and configure your Deployment options.
October 3, 2023
Additional Improvements
- Added a DAG Success alert so you can now set up an alert for successful completion events. See how to set up Astro alerts.
Bug Fixes
- Fixed a problem in the Astro UI where a warning about Deployment Health was displayed when a Workspace had zero Deployments.
September 26, 2023
Introducing the Astro API
The Astro API is currently in beta. See Astro API versioning and support.
You can now use the Astro API to create applications and scripts to programmatically interact with Astro. The Astro API is a standard REST API that includes endpoints for interacting with all key resources and components on Astro.
Using the Astro API, you can create robust and secure applications for managing Deployment resources, updating user permissions, and performing many other key Astro operations. To make your first API call, see Get started with the Astro API.
September 19, 2023
Manage Deployments programmatically using Deployment API tokens
Deployment API tokens replace Deployment API keys as the most secure and customizable way to manage Deployments programmatically. You can use Deployment API tokens to perform all of the same actions as a Deployment API key, including:
- Pushing code to a Deployment.
- Updating a Deployment's environment variables.
- Making requests to update your Deployment's Airflow environment using the Airflow REST API.
Unlike Deployment API keys, you can set an expiration date for Deployment API tokens and rotate them to better manage access to your Deployment. See Deployment API tokens to learn how to create and manage Deployment API tokens.
Deployment API tokens are a direct replacement for Deployment API keys. Therefore, Astronomer recommends always using Deployment API tokens over API keys. While you can still continue to use and manage existing Deployment API keys, Astronomer will soon require using Deployment API tokens.
After API tokens are generally available, Deployments with zero API keys will not show the API Keys tab and you will no longer be able to create Deployment API keys. If you want to continue using API keys, ensure that you always have at least one API key configured for the Deployment.
Additional improvements
- When you create a new Deployment, the Astro UI now presents new options and suggestions for running your first DAG.
- You can now retrieve a Workspace's ID from the Astro UI. To find a Workspace's ID, open the Workspace in the Astro UI and go to Workspace Settings > General.
September 12, 2023
Per-Deployment IAM workload identities on AWS
Astro Hybrid clusters on AWS now support per-Deployment IAM workload identities, meaning that you can now limit your trust policies to authorize only specific Deployments to your cloud resources.
This change required an automatic update to the cross-account role that Astro uses to manage clusters in your cloud. In addition to enabling per-Deployment IAM workload identities, this update also adds the following permissions to reduce the risk of partial deletions in your cloud:
{ "elasticloadbalancing:DescribeLoadBalancers", "elasticloadbalancing:DeleteLoadBalancer" }
For more information about this change, see Automatic updates coming to cross-account roles for Astro Hybrid on AWS.
To migrate from using cluster workload identities to Deployment workload identities:
-
In the AWS Management Console, go to the Identity and Access Management (IAM) dashboard. Identify all of your trust policies that specify your cluster workload identity. They should look similar to the following trust policy:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Principal": {
"AWS": [
"arn:aws:iam::<dataplane-AWS-account-ID>:role/AirflowS3Logs-<cluster-ID>"
]
},
"Action": "sts:AssumeRole"
}
]
} -
For each trust policy, add the workload identities for any Deployments that you want to access the related resource. To locate your Deployment workload identity, open the Deployment in the Astro UI and copy the Workload Identity from the Details page. Your trust policy should now look like the following:
{
"Version": "2012-10-17",
"Statement": [
{
"Effect": "Allow",
"Principal": {
"AWS": [
"arn:aws:iam::123456789876:role/AirflowS3Logs-cl6zcnlc641hr0voibivf21jh",
"arn:aws:iam::<dataplane-AWS-account-ID>:role/astro-<namespace>"
]
},
"Action": "sts:AssumeRole"
}
]
} -
For each Deployment that you specified in your trust policies, open the Deployment in the Astro UI and click Details, then click Edit Details. In the Workload Identity section, select the new Deployment identity from the dropdown list, then click Update. To avoid disruption to tasks, do not complete this step until you have added the Deployment workload identity to all of the trust policies it needs for access.
-
Upgrade to the latest Astro CLI release, which includes support for Per Deployment IAM Workload Identity.
-
After you've tested the policies with your Deployment workload identities, remove the cluster workload identity from your trust policies
Bug fixes
- Fixed an issue where Billing Admins could view task usage on the Usage page only for Workspaces that they belonged to. Now, Billing Admins can view usage for all Workspaces regardless of their Workspace role.
Additional improvements
- The Astro UI Usage page now shows task usage for deleted Deployments. If you're an Astro Hybrid Billing Admin, this means that task usage metrics now better reflect your billable usage.
- When you create a Deployment through the Astro UI and choose an Astro Runtime version, you can now select only the most recent supported patch for each major version of Astro Runtime.
- You can now filter task logs by log level or source from the DAGs page in the Astro UI.
September 6, 2023
View deploy history in the Astro UI
When you view a Deployment in the Astro UI, you can now open the Deploy History tab to view a table of all code deploys. The table shows who made deploys, when they made the deploys, and what Astro Runtime image they used for the deploy.
You can also now use the Astro CLI to specify an optional description for your deploys using the --description
flag. Deploy descriptions appear in the Deploy History table and are useful for telling other Workspace members why you made a deploy or what changes it contains. For more information, see View deploy history.
View rendered data figures in the Astro Cloud IDE
Python cells that generate figures using Matplotlib, Plotly, or any libraries that extend these tools now show the figures they render in the Astro Cloud IDE. When you run a Python cell, any figures it generates appear in a new Figures tab.
August 29, 2023
Changes to Workspace user roles
To take advantage of these new user roles programmatically, you must upgrade the Astro CLI to version 1.19 or later.
To increase granularity and better serve each user persona on Astro, Workspace roles have been updated with new names and permissions:
- The Workspace Author role is a new role for users who primarily write and deploy DAGs. Users with this role can push code changes, but they can't update Deployment or Airflow settings such as Airflow variables, Astro environment variables, or connections.
- The Workspace Admin role has been renamed to Workspace Owner. Users with this role are responsible for administrating membership to the Workspace.
- The Workspace Editor role has been renamed to Workspace Operator. In addition to pushing code changes, Workspace Editors can now also edit Airflow objects such as variables, connections, and XComs. Users with this role are responsible for managing the environments that DAGs run in.
- The Workspace Viewer role has been renamed to Workspace Member. Users with this role only need viewing permissions for a Deployment and don't have permissions to make any code or configuration changes.
For more information about these role changes, see User permissions reference and Enhanced Astro Workspace Roles for more granular permissions
Additional improvements
- The lifespan of the personal user access token you can retrieve from
cloud.astronomer.io/token
has been reduced from 24 hours to 1 hour. - The DAGs view of the Astro UI now shows your configured dependency edge labels in the graph view.
- The Astro UI now shows more detailed instructions for deploying code when you create a new Deployment.
- The Deployment Analytics page in the Astro UI has been renamed to Overview.
Bug fixes
- Fixed an issue where Deployments using the Kubernetes executor could not run DAGs with lower resource requests than the Default Pod Size. Minimum requests are now hard-coded and decoupled from default requests.
August 21, 2023
Additional improvements
- You can now configure task log forwarding to Datadog at the Deployment level.
- In the DAGs view of the Astro UI, you can now double click a task run node in the graph view to view the task run's logs and mapped tasks.
- The A50 worker type has been renamed to A60 to make it consistent in scale with other worker types.
- The max possible CPU quota and Memory quota for a Deployment running in a Hosted dedicated cluster has increased to 1600 vCPU/ 3200 GiB respectively.
August 15, 2023
Additional improvements
- You can now see how many Astro alerts you've configured for a DAG in the DAGs page of the Astro UI.
Bug fixes
- Fixed an issue where you couldn't run Astro Cloud IDE pipelines that included a Markdown cell.
- Fixed an issue where an Organization's SSO bypass link was formatted incorrectly in the Astro UI.
August 8, 2023
New Hosted worker type
You can now configure Deployments with the A50
machine type, which has 12 vCPU and 24 GiB. See Astro hosted resource reference.
Additional improvements
- When you create a new Astro Cloud IDE project, you can now specify whether you want the project to include an example pipeline.
- You can now access Organization-level settings in the Astro UI only through the Organization Settings link. Additionally, some Organization settings have been moved to the top level of navigation so that there is no longer a Settings menu.
- You can now use commas, apostrophes, and ampersands in Workspace and Organization names.
- The Workspace list view in the Astro UI has been redesigned so that Organization Owners can now edit and delete Workspaces directly from the list.
Bug fixes
- Fixed an issue where the Astro UI showed incorrect CPU and memory limits in bar charts on the Deployments list and Details page.
August 1, 2023
Hosted Deployments have DAG-only deploys enabled by default
New Astro Hosted Deployments now have DAG-only deploys enabled by default. When DAG-only deploys are enabled, some workflows for your Deployment, including image-based deploys, are different compared to when DAG-only deploys are disabled. For more information about how code and image deploys work when DAG-only deploys are enabled, see What happens during a project deploy. To disable DAG-only deploys, see Enable/ disable DAG-only deploys on a Deployment.
Teams now have Organization-level roles
All Teams on Astro now have an Organization role. Existing Teams have been given the Organization Member role, which doesn't result in any additional automatic permissions.
Coupled with SCIM user groups, you can now manage your Organization Owners and Billing Admins from your identity provider. See Manage teams for more information.
Additional improvements
- The Astro UI now shows how many Workspaces, DAGs, clusters, and Astro Cloud IDE projects you have in the left sidebar.
- You can now create Deployments in standard clusters hosted in
europe-west2
on GCP andeu-central-1
on AWS. - The default metadata database instance type for new Deployments on GCP clusters in Astro Hybrid has been reduced to
Small General Purpose
with 2 vCPUs and 8GiB. See GCP Hybrid cluster settings.
Bug fixes
-
Because some regions don't support specific machine types that Astro Hosted uses, you can no longer create Hosted dedicated clusters in the following AWS regions:
af-south-1
ap-east-1
ap-northeast-2
ap-northeast-3
ca-central-1
eu-south-1
eu-west-3
me-south-1
July 25, 2023
Additional improvements
- The templates for Astro alert messages have been updated to include more information about the Deployment that the alert was triggered in, including a link to the DAG that triggered the alert.
Bug fixes
- Fixed an issue where Azure AD single sign-on (SSO) connections were incorrectly labeled as SAML connections in the Astro UI.
July 18, 2023
Configure default Pod sizes
You can now configure the default minimum CPU and memory for tasks that you run with the Kubernetes executor or KubernetesPodOperator. If you don't specify CPU or memory in a task definition, Astro runs the task in a Pod that uses your default resource configurations. Configure default minimum resources to ensure that tasks always have enough CPU and memory to run successfully. See Deployment settings.
New regions available on Astro Hosted
You can now create Hosted dedicated clusters in the following regions:
-
AWS
af-south-1
- Africa (Cape Town)ap-east-1
- Asia Pacific (Hong Kong)ap-northeast-1
- Asia Pacific (Tokyo)ap-northeast-2
- Asia Pacific (Seoul)ap-northeast-3
- Asia Pacific (Osaka)ap-southeast-1
- Asia Pacific (Singapore)ap-southeast-2
- Asia Pacific (Sydney)ap-south-1
- Asia Pacific (Mumbai)ca-central-1
- Canada (Central)eu-central-1
- Europe (Frankfurt)eu-south-1
- Europe (Milan)eu-west-1
- Europe (Ireland)eu-west-2
- Europe (London)eu-west-3
- Europe (Paris)me-south-1
- Middle East (Bahrain)sa-east-1
- South America (São Paulo)us-east-1
- US East (N. Virginia)us-east-2
- US East (Ohio)us-west-1
- US West (N. California)us-west-2
- US West (Oregon)
-
GCP
asia-east1
- Taiwan, Asiaasia-northeast1
- Tokyo, Asiaasia-northeast2
- Osaka, Asiaasia-northeast3
- Seoul, Asiaasia-south1
- Mumbai, Asiaasia-south2
- Delhi, Asiaasia-southeast1
- Singapore, Asiaasia-southeast2
- Jakarta, Asiaaustralia-southeast1
- Sydney, Australiaaustralia-southeast2
- Melbourne, Australiaeurope-central2
- Warsaw, Europeeurope-north1
- Finland, Europeeurope-southwest1
- Madrid, Europeeurope-west1
- Belgium, Europeeurope-west2
- England, Europeeurope-west3
- Frankfurt, Europeeurope-west4
- Netherlands, Europeeurope-west6
- Zurich, Europeeurope-west8
- Milan, Europeeurope-west9
- Paris, Europenorthamerica-northeast1
- Montreal, North Americanorthamerica-northeast2
- Toronto, North Americasouthamerica-east1
- Sau Paolo, South Americasouthamerica-west1
- Santiago, South Americaus-central1
- Iowa, North Americaus-east1
- South Carolina, North Americaus-east4
- Virginia, North Americaus-east5
- Columbus, North Americaus-south1
- Dallas, North Americaus-west1
- Oregon, North Americaus-west2
- Los Angeles, North Americaus-west3
- Salt Lake City, North Americaus-west4
- Nevada, North America
See Astro Hosted resource reference for all available configurations.
Bug fixes
- You can no longer create Hybrid GCP clusters in
eu-north-1
.
July 11, 2023
Send Astro alerts to email
You can now send Astro alerts to multiple email addresses. Sending Astro alerts to email requires no configuration outside of Astro, which makes it a quick option to improve your alerting infrastructure. See Configure Astro alerts for setup steps.
Configure SCIM provisioning for Azure AD
If your Organization uses Azure for single sign-on (SSO), you can now set up SCIM provisioning for Astro. SCIM provisioning simplifies user management by allowing you to add and remove Astro users from Okta based on your existing user groups. See Set up SCIM provisioning for more information.
Additional improvements
- On Astro Hosted deployments, the
astronomer_monitoring_dag
has been paused for all image-based Deployments and removed entirely from all Deployments with DAG deploys enabled. It has been replaced with an implementation that allows workers on Deployments to fully scale to 0.
July 5, 2023
Configure SCIM provisioning for Okta
If your Organization uses Okta for single sign-on (SSO), you can now set up SCIM provisioning for Astro. SCIM provisioning simplifies user management by allowing you to add and remove Astro users from Okta based on your existing user groups. See Set up SCIM provisioning for more information.
See pricing estimate when creating a Deployment
The Deployment creation page in the Astro UI has been reorganized to make it easier to focus on specific configurations for your Deployment. Each configuration is now collapsible and includes guidance for different environment sizes. Additionally, the page now shows cost estimates for a Deployment before you create it.
Additional improvements
- You can now configure Deployments with the
A40
machine type, which has 8 vCPU and 16 GiB. See Astro hosted resource reference. - You can now access your Organization settings from a Workspace by clicking the name of your Organization/ Workspace.
- On Astro Hybrid, the default Azure DB instance is now
Standard D2ds_v4
. - The Deployment creation screen for Astro Hybrid has received several updates that were previously only available on Astro Hosted.
Bug fixes
- Fixed an issue where DAGs that used the KubernetesPodOperator and had tasks with
in_cluster=False
could not be parsed.
June 27, 2023
Support for dedicated clusters on Azure
You can now create a dedicated cluster in the following Azure regions:
australiaeast
eastus2
northeurope
westeurope
uswest2
See Astro Hosted resource reference for more information.
Additional improvements
- The Astro UI now shows how many Workspaces each Team belongs to in Settings > Access Management > Teams.
- You can now create dedicated clusters in
us-west1
on GCP.
June 20, 2023
Additional improvements
- You can now add a new Astro user to Workspaces before the user has accepted their invite.
- If you have Organization Owner permissions, you can now add a user to a Workspace even if the user hasn't been added to your Organization. Users added to Workspaces this way are automatically added to your Organization as an Organization Member.
- The Astro UI now shows your Team IDs in Settings > Access Management > Teams. Use Team IDs to add Teams to Workspaces using the Astro CLI.
- A Team's Updated At and Updated By values are now updated when you change the Team's permissions in a Workspace or Organization.
Bug fixes
- Fixed an issue where a Workspace descriptions were incorrectly required when creating a new Workspace through the Astro CLI.
June 13, 2023
Manage billing and track usage for Astro Hosted
Use the new Billing page in the Astro UI to see both high-level and detailed metrics about your spend in Astro Hosted. You can also use this page to configure your billing details and view invoices. See Manage billing for more details.
New cell type for using Airflow operators in the Astro Cloud IDE
You can now use any Airflow operator available on the Astronomer Registry in your Astro Cloud IDE pipeline. Operator cells apply formatting and checks for parameter inputs, making it easy to configure operators as part of your pipeline. See Use Airflow operators in the Astro Cloud IDE.
Additionally, you can configure custom cells to use your team's custom operators in a pipeline. See Create custom operator cells.
IMDSv2 is now enforced on AWS clusters
If your DAGs assume IAM roles to directly access metadata on your cluster using IMDSv1, this change can result in DAG run failures. Upgrade your DAGs to use IMDSv2 for all cluster metadata requests.
See Use IMDSv2 for more information.
Astronomer now enforces IMDSv2 on all AWS clusters. Any requests for resources on your clusters are now session-based and muse include a token in the request body.
Additional improvements
- Trial Deployments now have DAG-only deploys enabled by default.
- The Astro UI now shows your Organization Short Name and Astro SAML Connection Name in the Astro UI.
- You can now view mapped tasks from the DAGs page in the Astro UI.
Bug fixes
- Fixed an issue where worker node pools in Hosted dedicated clusters on Azure were not being updated correctly.
- Fixed an issue the Astro UI would reset a Deployment's Min Worker Count from 0 to 1 after you edited the Deployment in any way.
June 6, 2023
Track user actions and ensure compliance with audit logs
You can now export audit logs from the Astro UI to view all actions taken in your Organization over a given time period. See Export audit logs for setup steps.
Additional improvements
- You can now configure Hybrid GCP clusters with additional Memory Optimized and Compute Optimized Cloud SQL instance types. See Supported Cloud SQL instance types.
May 30, 2023
Manage permissions for groups of users with Teams
Configure Teams from the Astro UI to manage the permissions for many users across Workspaces from a single page. Teams are a group of users in an Organization that you grant the same Workspace permissions, without needing to define them individually.
See Make a Team for setup steps.
Bug fixes
- In Astro Hosted, an irrelevant AWS external ID info page has been removed from the Astro UI.
- Fixed an issue where DAG-only deploys could be unreliable due to the deploy process not requesting enough resources in the cluster.
May 23, 2023
Introducing Astro Hosted and Hybrid
Astro Hosted is a new way to run Airflow on Astronomer's cloud. On Astro Hosted, Airflow environments are managed and hosted entirely by Astronomer, enabling you to shift your focus from infrastructure to data.
For more information about how Astro Hosted works, see the Architecture overview.
If you're already an Astro user and your Deployments run in your company's own cloud, you're using Astro Hybrid. This version of Astro was formerly known as Astro - Bring Your Own Cloud.
To see whether you're an Astro Hybrid user, open your Organization in the Astro UI and go to Settings > General. Your version of Astro is listed under Product Type.
See Documentation refactor for Astro Hybrid to learn how the documentation has changed for current Astro Hybrid users.
Configure default Kubernetes Pods on Astro Hosted
One of the biggest risks of running the Kubernetes executor or KubernetesPodOperator is that your tasks can accidentally request more resources than expected, which can drive up costs. To limit this risk, you can now configure default and maximum Pod resources from the Astro UI. If a task tries to request Pod resources that are more than your configured limits, the task fails.
See Configure Kubernetes Pod resources for setup steps.
Documentation refactor for Astro Hybrid
The following updates have been made to documentation to accommodate new Astro Hosted information:
- Astro Hybrid documentation now has a dedicated menu under "Administration" that contains all docs related to Hybrid installation and cluster management. See Astro Hybrid overview.
- All other docs now assume Astro Hosted by default. If a feature is functionally different between Hosted and Hybrid, the documentation for that feature will include a note about how that setup differs for Hybrid. Look for the blue Alternative Astro Hybrid setup notes throughout documentation.
May 16, 2023
Automate Organization management with Organization API tokens
You can now create Organization API tokens to automate key actions across your Organization and all of the Workspaces in it. You can customize the role and expiration date of the token to give it the minimum required permissions for the task it completes. Some common actions that you can automate with Organization API token are:
- Creating Workspaces.
- Inviting users to an Organization or Workspace.
- Creating and updating Deployments using a Deployment file.
- Exporting audit logs.
- Gathering metadata about Deployments using the Airflow REST API.
- Completing any of the actions you can complete with a Workspace API token or Deployment API key across all Deployments in your Organization.
See Manage Organization API tokens for more information.
May 2, 2023
Receive Astro alerts on Slack or PagerDuty
Astro alerts are a new way to be notified when your DAGs aren't running as expected. Unlike Airflow callbacks and SLAs, Astro alerts require no changes to DAG code and integrate with Slack and PagerDuty.
You can set an alert on any DAG to be notified when the DAG fails or when a task takes longer to run than expected. See Astro alerts for configuration steps.
Bug fixes
- Fixed an issue where SSO configurations made through Astronomer support could be overridden by updating the SSO configuration through the Astro UI.
April 26, 2023
Improved log viewing in the Astro UI
The Deployment Logs page in the Astro UI now shows logs for your Deployment's workers, schedulers, triggerers, and webserver. Additionally, you can now view up to the last 10,000 logs emitted by your Deployment from the Astro UI.
To make it easier to parse this larger log volume, the Logs page now lets you filter by log type, date, and keyword. See View logs for more information.
April 18, 2023
Self-service configuration for single sign-on (SSO) connections
You can now configure SSO connections directly from the Astro UI without assistance from Astronomer support. Use the Authentication page to configure different authentication environments for your Organization by creating and managing multiple SSO connections and domains.
To review the new process for creating SSO connections, see Set up authentication and SSO. To create new managed domains to map to your SSO connections, see Manage domains.
April 11, 2023
Additional improvements
- The node type for running Airflow system components on GCP clusters has been reduced from
n2-standard-4
toe2-standard-4
. - To optimize infrastructure costs for running the Kubernetes executor, Kubernetes executor worker Pods from different Deployments can now run on the same worker node. This occurs only when the Deployments are hosted in the same cluster and use the same worker node instance type.
April 4, 2023
Preview Deployments
You can now create preview Deployments from feature branches in your Git repository. Use a preview Deployment template or GitHub Actions template to configure your Astro pipelines to:
- Create the preview Deployment when you create a new branch.
- Deploy code changes to Astro when you make updates in the branch.
- Delete the preview Deployment when you delete the branch.
- Deploy your changes to your base Deployment after you merge your changes into your main branch.
Additional improvements
- Added the ability to enforce CI/CD deploys. You can now configure your Deployment to only accept code deploys if they are triggered by a Deployment API key or Workspace token.
- When you create a new cell in the Astro Cloud IDE, the editor auto-scrolls to your new cell and selects it.
Bug fixes
- Fixed a bug where the UI passed the wrong cluster type.
- Fixed an issue where the Deployment status shows as 'deploying' when KPOs are running.
March 28, 2023
New GCP node instance types available
You can now use the following node instance types for worker nodes in GCP clusters:
e2-standard-32
e2-highcpu-32
n2-standard-32
n2-standard-48
n2-standard-64
n2-highmem-32
n2-highmem-48
n2-highmem-64
n2-highcpu-32
n2-highcpu-48
n2-highcpu-64
For a list of all instance types available for GCP, see Supported worker node pool instance types.
Additional improvements
- You can now use
db.m6g
anddb.r6g
RDS instance types on AWS clusters. - The default RDS instance type for new AWS clusters has been reduced from
db.r5.large
todb.m6g.large
- The default CIDR range for new AWS clusters has been reduced from /19 to /20.
- You can now submit a Request type in the Astro UI support form. When you choose a request type, the form updates to help you submit the most relevant information for your support request.
- You can no longer delete a Workspace if there are any Astro Cloud IDE projects still in the Workspace.
- Organization role permissions have changed so that only Organization Owners can create Workspaces.
Bug fixes
- Fixed an issue where you could set a Deployment's scheduler resources to less than 5 AU.
March 21, 2023
Automate Workspace and Deployment actions using Workspace API tokens
Use Workspace API tokens to automate Workspace actions, such as adding users to a Workspace and creating new Deployments, or for processes that a Deployment API key can automate. You can customize the role and expiration date of the token to give it the minimum required permissions for the task it completes.
To create and use Workspace API tokens, see Workspace API tokens.
Additional improvements
-
In the Astro Cloud IDE, you can now specify the output table for a Warehouse SQL cell using both literal and Python expressions. See Create a SQL cell.
-
Port 80 is no longer used for certificate management on the data plane.
-
To switch Organizations in the Astro UI, you now use the Switch Organization button next to your Organization's name.
March 15, 2023
Run the Kubernetes executor in Astro
You can now configure your Deployments to use the Kubernetes executor for executing tasks. Using the Kubernetes executor, you can:
- Ensure that tasks running longer than 24 hours are not interrupted when your team deploys code.
- Run tasks with different version dependencies in the same Astro project.
- Request specific amounts of CPU and memory for individual tasks.
- Automatically down your resources when no tasks are running.
The Kubernetes executor runs each task in its own Kubernetes Pod instead of in shared Celery workers. Astronomer fully manages the infrastructure required to run the executor and automatically spins Pods up and down for each of your task runs. This executor is a good fit for teams that want fine-grained control over the execution environment for each of their tasks.
To learn whether the Kubernetes executor works for your use case, see Choose an executor. To configure the Kubernetes executor for a task or Deployment, see Configure the Kubernetes executor.
Simplified Organization management in the Astro UI
The Astro UI has been redesigned so that Organization settings tabs are now available in the left menu. Use this new menu to switch between pages as you can for Workspace settings.
While most tabs were migrated directly to the left menu with the same name, some pages have been renamed and moved:
- Formerly located in Overview, your Workspace list is now available in Workspaces.
- Formerly located in the People tab, Organization user management settings are now in Settings > Access Management.
- Formerly located in the Settings tab, general Organization settings are now in Settings > General.
New Astro Cloud IDE integration with GitLab
You can now configure a GitLab repository in your Astro Cloud IDE project. Configuring a GitLab repository allows you to commit your pipelines and deploy them to Astro directly from the Astro Cloud IDE. See Deploy a project from a Git repository to Astro.
Additional improvements
- Clusters on an Astro - Hosted installation no longer retain Airflow logs which are older than 90 days.
- The Data Plane System node pool instance type on GCP clusters has been reduced from
n2-standard-4
ton2-standard-2
.
March 7, 2023
Get expert advice on Astro and Airflow in office hours
Office hours are a new way for Astro customers to meet with the Astronomer Data Engineering team. In an office hour meeting, you can ask questions, make feature requests, or get expert advice for your data pipelines.
You can now schedule a 30-minute office hour meeting in the Help menu next to your user profile in the Astro UI.
For more information, see Book office hours in the Astro UI.
Additional improvements
- The node pool instance type used for Astro system components on GCP clusters has been reduced from
n2-standard-4
ton2-standard-2
. - DAGs generated by the Astro Cloud IDE now use UTC instead of your current timezone as the default timezone for scheduling DAG runs.
Bug fixes
- Fixed an issue in the Astro Cloud IDE where you could not update a pipeline that was configured with an invalid cyclic dependency chain.
- Fixed an issue where deploying an Astro project with a custom Docker image tag resulted in the Deployment always having the Deploying status in the Astro UI.
- Fixed an issue where worker Pods on Azure clusters were sometimes unable to scale because there was no prioritization for starting up essential scheduling Pods.
March 1, 2023
Astro no longer requires administrator access on AWS
Astro no longer requires administrator permissions for its dedicated AWS account. Instead, Astro now assumes a cross-account IAM role with the minimum necessary permissions for running and managing clusters. See Install Astro on AWS for more information.
IdP-initiated logins through the Okta dashboard
If your Organization uses Okta as your Astro identity provider, you can now log in to Astro directly from your Okta Apps dashboard. If you've been authenticated by Okta, you no longer need to be authenticated by Astro when you access it through your dashboard.
Additional improvements
- Ingress to the Kubernetes API on Google Cloud Platform (GCP) and Azure clusters is now limited to Astro control plane IPs. This change will be implemented on all clusters in the coming weeks.
Bug fixes
-
To protect the functionality of Astro monitoring services, you can no longer override the values of the following environment variables:
-
AIRFLOW__METRICS__STATSD_ON
-
AIRFLOW__METRICS__STATSD_HOST
-
AIRFLOW__METRICS__STATSD_PORT
-
AIRFLOW__METRICS__STATSD_ALLOW_LIST
-
AIRFLOW__METRICS__STATSD_STATSD_CUSTOM_CLIENT_PATH
-
AIRFLOW__METRICS__STATSD_PREFIX
You can still set new values for these variables, but the values will be automatically overwritten in the Astro data plane. See Platform variables.
-
-
Fixed an issue where a user could provision multiple accounts when their login email address included differently cased characters.
February 21, 2023
New identity-first authentication model
Astro has migrated to an identity-first authentication model. Users now authenticate to the Astro platform instead of individual Organizations, and Organizations can set permissions for how users can modify and access resources. This model prioritizes identity verification and enforces authentication policies for user email domains.
For all users logging in to Astro, this migration has the following effects:
- Instead of being redirected to separate login pages for each Organization, all Astro users log in through a universal login page.
- Users belonging to multiple Organizations no longer have to log in again when switching Organizations.
- Users no longer need to enter their email on a separate page before they log in to the Astro UI.
- If your Organization enforces single sign-on (SSO), users can now authenticate to Astro with a username and password when your email domain doesn't enforce SSO.
For Organization Owners, this migration has the following additional effects:
- You can now use an SSO bypass link to log in to Astro if your SSO connection is disrupted.
- Your Organization now has a list of owned email domains, and any users logging into Astro with one of those domains will be redirected to your configured identity provider.
To configure authentication behavior, see Configure SSO.
New Hosted regions available
You can now create clusters in the following regions on an Astro - Hosted installation.
-
AWS
ap-northeast-1
ap-southeast-2
eu-central-1
eu-west-1
us-east-1
us-west-2
-
Google Cloud
asia-northeast1
australia-southeast1
europe-west1
europe-west2
us-central1
us-east4
-
Microsoft Azure
australiaeast
japaneast
northeurope
westeurope
eastus2
westus2
Additional improvements
The default CIDR ranges for new GCP clusters have been reduced. The following are the new CIDR ranges:
- Subnet CIDR:
172.20.0.0/22
- Pod CIDR:
172.21.0.0/19
- Service Address CIDR:
172.22.0.0/22
- Service VPC Peering:
172.23.0.0/20
Bug fixes
In the Astro UI, when using Compare on the Lineage Graph page, you can now compare shorter run lengths.
February 14, 2023
Authorize Workspaces to clusters
You can now keep teams and projects isolated by authorizing Workspaces to specific clusters. Use this feature to better manage cloud resources by ensuring that only authorized Deployments are running on specific clusters.
New Deployment health statuses and information in the Astro UI
The Astro UI now includes three additional Deployment health statuses that you might see when creating or pushing code to a Deployment.
- The Creating status indicates that Astro is still provisioning the resources for the Deployment.
- The Deploying status indicates that a code deploy is in progress. Hover over the status indicator to view specific information about the deploy, including whether it was an image deploy or a DAG-only deploy.
- The Unknown status indicates that Deployment status can't be determined.
Additionally, the Deployment information page in the Astro UI now includes fields for Docker Image and DAG Bundle Version that show unique timestamps and tags based on your latest code deploy. Use this information as the source of truth for which version of your code is currently running on the Deployment.
View OpenLineage facets for lineage job runs
OpenLineage facets are JSON objects that provide additional context about a given job run. By default, a job run includes facets that show what kind of job was completed, whether the job run was successful, and who owns the job.
You can now view all available facets for a job run, including custom facets, by opening the job run's Lineage Graph and then selecting the Info tab. You can check the status of your facets, including whether they are correctly formatted, so that you can resolve potential issues in your data pipelines. See View metrics for a specific run or dataset.
Additional improvements
- You can now create AWS clusters in
ap-northeast-1
andap-southeast-2
on an Astro - Hosted installation. - You can now create GCP clusters in
australia-southeast1
on an Astro - Hosted installation.
Security fixes
- Fixed CVE-2023-0286.
February 7, 2023
Additional improvements
- The instructions on the welcome page for new Astro users who are not yet part of an Organization have been improved to make getting started easier.
Bug fixes
- Removed nonfunctioning date filtering functionality from the Lineage UI.
- Fixed an issue where triggering an image deploy from an older version of the Astro CLI could unintentionally turn off DAG deploys on a Deployment.
January 31, 2023
Bug fixes
- When you select Mark Success or Clear for Deployment task actions in the Airflow UI, you are now correctly redirected to the DAG Tree view instead of the DAGs homepage.
- Fixed CVE-2022-41721.
January 24, 2023
New Workspace Home page
When you select a Workspace in the Astro UI, the Home page now appears first. On this page, you can:
- Check the status of your Deployments.
- Quickly access your most recently viewed Deployments and Cloud IDE projects.
- View release notes for all Astro products.
See Introducing Astro’s New Workspace Homepage for more information.
Additional improvements
- Ingress to the Airflow UI and API on Astro clusters is now limited to control plane IPs. This change will be implemented on all clusters in the coming weeks.
- You can now request custom tags for your AWS clusters by submitting a support request to Astronomer support. You can view your cluster tags in the Astro UI by selecting Clusters, selecting a cluster, and then clicking the Details tab.
- You can now create new clusters in France Central for Bring Your Own Cloud installations of Astro on Azure.
- Improved the speed of DAGs appearing in the Airflow after completing a DAG-only deploy.
Bug fixes
- Fixed CVE-2022-48195.
January 18, 2023
Bug fixes
- Fixed an issue with Google Cloud Platform (GCP) clusters where the metadata database for a Deployment could persist after the Deployment was deleted.
January 10, 2023
New Astro Cloud IDE cell types
To simplify the creation of new tasks, the following new cell types are now available in the Astro Cloud IDE:
- SQL: Run a SQL query against an existing database connection and save the query results in an XCom file for use by other cells. Use this cell type to run smaller queries and store the results in Airflow for quick access by other cells.
- Warehouse SQL: Run a SQL query against an existing database connection and store the query results in your data warehouse. Use this cell type for data operations that require more storage and reliability.
- Markdown: Add inline Markdown comments to your generated DAG code. Use this cell type to document code decisions and to make it easier for team members to collaborate on shared pipelines.
For more information about a specific cell type, see Run SQL in the Astro Cloud IDE, Run Python in the Cloud IDE, and Add documentation to an Astro Cloud IDE pipeline.
Additional improvements
- To reduce the time it takes for Airflow to parse new DAG files, the default value for
AIRFLOW__SCHEDULER__DAG_DIR_LIST_INTERVAL
has been reduced from 5 minutes to 30 seconds for all Deployments regardless of Runtime version. For most users, this means that you will see new DAGs appear in the Airflow UI faster. - In the Astro UI, a banner now appears if there is an incident reported on the Astro status page.
Bug fixes
- Sorting the Organization Role column in the People tab of the Astro UI now works as expected.
- Fixed an issue where lineage groups would occasionally not collapse as expected in the Lineage Graph view.
December 20, 2022
Additional improvements
-
You can now configure OneLogin and Ping Identity as identity providers on Astro.
-
Workspace Members can now view Workspace settings in the Astro UI.
-
Node groups that are collapsed in the lineage graph in the Astro UI now show only the total number of connected Jobs and Datasets, instead of listing each job and dataset. This makes the lineage graph easier to navigate.
Bug fixes
- Fixed an issue where the lineage UI did not show dataset metrics when a dataset had no column-level metrics.
- Fixed an issue where some instances of a dataset's name were inconsistent in the lineage UI.
December 13, 2022
Improvements to the Cloud IDE
The Cloud IDE includes several new features which improve DAG authoring and testing:
- There is a new Commit button in the Astro UI that is separate from the Configuring GitHub menu.
- The default CI/CD pipeline included in the Cloud IDE project supports DAG-only deploys. Deploying DAG changes to Astro using the CI/CD pipeline is now significantly faster.
- The Configure GitHub menu in the Astro UI now includes a Clone Repo settings menu. Enabling this option makes other files in your GitHub repository, such as helper functions in the
include
folder of your project, accessible when you run DAGs in the Cloud IDE. - You can now explicitly mark upstream dependencies for a task cell from the cell's configuration menu.
For more information about configuring GitHub and deploying code with the Cloud IDE, see Deploy a project from the Cloud IDE to Astro.
Support for n2 worker types on GCP
You can now configure worker queues with the following n2
worker types on Google Cloud Platform (GCP) clusters:
n2-standard-4
n2-standard-8
n2-standard-16
n2-highmem-4
n2-highmem-8
n2-highmem-16
n2-highcpu-4
n2-highcpu-8
n2-highcpu-16
For more information about these worker types, see N2 machine series. For a list of all worker types available on GCP, see Worker node size resource reference.
Additional improvements
- In the Clusters tab of the Astro UI, you can now click a cluster entry to see details about the cluster configuration, including which Worker Types are enabled for the cluster.
- The Deployment details page in the Astro UI now includes an ID pane. A Deployment ID is required when you deploy code using a CI/CD process.
- The OpenLineage URL for your Organization is now available on the Settings page in the Astro UI. An OpenLineage URL is required to integrate data lineage from some external systems.
- Workspaces are now sorted alphabetically in the Astro UI.
- In Astro CLI version 1.8.0 or later, running
astro deploy
with an empty or missingdags
folder does not erase or override existing DAGs. Instead, the directory is excluded from the build and push process to Astro. This lets you manage your DAGs and project files in separate repositories when using DAG-only deploys.
Bug fixes
- Fixed an issue where Astro temporarily stored DAGs for DAG-only deploys in a new directory named
/usr/local/airflow/dags/current
, which could cause import errors in user code. - Fixed an issue where task runs triggered in the Cloud IDE did not have access to project environment variables.
- Fixed an issue where Deployment metrics for memory usage were not always accurate.
November 15, 2022
Additional improvements
- In the Astro UI, the People page now shows the IDs of users belonging to your Organization.
- In the Astro UI, the Deployments page now shows the user or API key that most recently updated each Deployment and when they updated it.
Bug fixes
- Availability zone (AZ) rebalancing has been disabled for worker node pools on AWS clusters. This change should result in fewer zombie tasks and less volatility across workers. AZ rebalancing is enabled for other system components on Astro.
- The Updated at field for a transferred Deployment now displays the correct time.
astro deploy --dags
now handles deferrable tasks correctly.
November 8, 2022
Deploy only DAGs with astro deploy -—dags
Using Astro CLI 1.7, you can run astro deploy -—dags
to push only the dags
directory of your Astro project to a Deployment on Astro. This is an additional option to astro deploy
that makes for a faster development experience and gives you more flexibility in how you configure CI/CD processes.
For more information, see Astro CLI 1.7 or Deploy DAGs only. For example CI/CD workflows with this feature enabled, see CI/CD.
Improved data lineage interface
The Lineage tab has new features and is better integrated into the Astro UI.
Specifically, the tab includes the following improvements:
- The process for comparing runs uses a simpler interface and provides more information about the runs you're comparing. See Compare lineage graphs from previous runs.
- Names for UI elements have been updated to more clearly represent Airflow resources. For example, jobs is now runs, and the Explore tab is now Runs.
- Lineage graphs include new colors and animations to show the flow of data as it moves between runs and datasets.
Transfer a Deployment
You can now transfer a Deployment from one Workspace to another in your Organization. This feature is helpful if you need to change the group of users that have access to a Deployment, or if you create a Deployment in the wrong Workspace.
See Transfer a Deployment to another Workspace.
Additional improvements
- The Kubernetes API is no longer exposed to the public internet on AWS data planes. The allowlist is limited to control plane IPs. New clusters will be created with this configuration, while all existing clusters will be updated by end of next week.
November 1, 2022
Introducing the Astro Cloud IDE, a new Airflow development experience
Astronomer is excited to introduce the Astro Cloud IDE, which is a notebook-inspired development environment for writing, running, and deploying data pipelines. Now you can develop an entire Airflow project, including DAGs, dependencies, and connections entirely within the Astro UI.
The Astro Cloud IDE was created with the following objectives:
- Configuring Airflow shouldn't be a barrier to running Airflow.
- Passing data between tasks should be seamless regardless of what language is used to write the task.
- Data pipelines should be quick to deploy and easy to test with CI/CD.
Most importantly, the Astro Cloud IDE was developed to make it easier for new Airflow users to get started and to provide experienced users with a robust development environment.
To create your first project in the Astro Cloud IDE, see the Cloud IDE quickstart. To deploy your project to Astro, see Deploy your Cloud IDE project to Astro.
Additional improvements
- In the Astro UI, cluster selection menus are now alphabetized.
Bug fixes
- Fixed an issue where the KubernetesPodOperator was not aware of available ephemeral storage in
m5d
andm6id
worker nodes. This issue resulted in Pods being evicted to free up storage even when there was enough available storage for tasks. - Fixed an issue in the Astro UI where you could select a worker type before selecting a cluster when creating a Deployment.
- Fixed an issue where Deployments on Runtime 5.0.10 and earlier showed a nonfunctional Configuration tab in the Airflow UI.
- Fixed CVE-2022-32149.
October 25, 2022
Additional improvements
- In the Astro UI, you can now view a cluster's external IP addresses in the Clusters tab.
Bug fixes
- Fixed an issue where some Deployments were running tasks after being deleted.
October 18, 2022
Additional improvements
- In the Astro UI, Access has been moved from the left menu to a tab on the Workspace Settings page.
- In the Astro UI, Workspace Settings in the left menu is now available to all Workspace members.
New Azure regions
You can now create an Astro cluster on Azure in the following regions:
japaneast
southafricanorth
southcentralus
October 11, 2022
Additional improvements
- New worker node pools on Azure and Google Cloud Platform (GCP) clusters can now scale to zero. When you set your minimum worker count to 0, you don't incur costs for enabling a new worker type for your cluster until it's used in a Deployment.
Bug fixes
- Fixed an issue where worker queues with a minimum worker count of zero would appear with a minimum worker count of one in the Astro UI.
October 4, 2022
New permissions boundary for managed AWS Accounts
The operational roles that Astronomer assumes on dedicated customer AWS accounts now have new permissions boundaries that limit the roles to a subset of their permissions. The remote management role is now limited to the following actions across all contexts:
autoscaling:*
cloudformation:*
cloudwatch:*
ec2:*
ecr:*
eks:*
elasticloadbalancing:*
iam:*OpenID*
kms:DescribeKey
lambda:*
logs:*
route53:AssociateVPCWithHostedZone
s3:*
secretsmanager:*
servicequotas:*
ssm:*
tag:*
These permissions might change in the future to enable new Astro features or to refine permissions for specific contexts.
Additional improvements
- Users with the required permissions can now access a Configuration tab in the Admin menu of the Airflow UI. This page no longer shows sensitive values in plain-text and can be used to verify all configurations running on your Deployment.
- In the Astro UI, the maximum time for Deployment metrics has been extended from 24 hours to 7 days.
- The Deployment metrics overview now shows metrics for the
default
worker queue instead of an aggregate of all worker queues. Improved worker queue metrics coming soon.
Bug fixes
- Added the global environment variable
AIRFLOW__LOGGING__DAG_PROCESSOR_LOG_TARGET=stdout
so that a scheduler's logs don't overcrowd its local storage - Removed misleading maximum CPU and memory lines from Deployment metric graphs
September 28, 2022
Additional improvements
- All worker queue configurations in the Worker Queues tab of the Astro UI now have tooltips.
- The Worker CPU and Worker Memory metrics in the Analytics tab of the Astro UI now show metrics only for the default worker queue instead of an average across queues. Improved worker queue metrics coming soon.
Bug fixes
- Values in the Organization Settings page no longer overlap with other UI elements.
- Organization Owners can now push code to a Deployment even if they aren't explicit members of the Deployment's Workspace.
September 21, 2022
A simpler Deployment page
All of a Deployment's configurations, including analytics, API keys, environment variables, and resource configurations, are now organized as tabs within the Deployment's page in the Astro UI.
This new UI moves the Analytics and Logs from the left sidebar to the main Deployment page so that you no longer have to filter those views separately by Deployment. The left sidebar now exclusively contains Workspace-level menus.
New Account Dashboard
You can now access your Account Dashboard to manage your user account settings and find links to helpful resources. Access this page by going to account.astronomer.io
in your browser or by clicking Profile > Manage your Astro Account in the Astro UI. You must be authenticated to Astro.
Additional improvements
- You can now use the
m6id
worker node type series for Deployments on AWS clusters. This worker type is general purpose and includes significant storage as well as up to 15% better performance compared tom5d
nodes. For more information, see Worker instance types. - New worker node pools on Amazon Web Services (AWS) clusters can now scale to zero. This means that enabling a new worker type for your cluster does not cost you until it's used in a Deployment.
Bug fixes
- Fixed an issue where the Astro UI Deployment metrics showed a maximum worker CPU and memory that was inconsistent with your configured worker queues.
September 14, 2022
Additional improvements
- When you create a new worker queue, the default worker type in your cluster is now pre-selected in the Worker Type list.
- You can now configure multiple instances of the same identity provider (IdP). See Configure an identity provider.
- You can now expand and collapse the Workspace menu in the Astro UI.
Bug fixes
- Fixed an issue where you could not open the Airflow UI from a Deployment.
August 31, 2022
Export Deployment metrics to Datadog
You can now export over 40 Airflow metrics related to the state of your Astro Deployment to Datadog by adding a Datadog API key to the Deployment. Metrics include task successes, DAG processing time, frequency of import errors, and more.
For organizations already using the observability service, this integration allows your team to standardize on tooling and gain a more granular view of Deployment metrics in a single place. Once the integration is configured, Astro automatically exports all available metrics to Datadog. For a complete list of supported metrics, see Data Collected.
To learn more, see Export Airflow metrics to Datadog.
Additional improvements
- The Astro UI now automatically ensures that worker queue names are valid as you type in real time.
- The number of times that a user can enter the wrong credentials for Astro before being locked out has been reduced from 10 to 6.
- You can now configure worker queues to have a minimum Worker count of 0 workers. Note that depending on your cloud provider and Deployment configurations, some Deployments still might not be able to scale to 0 workers.
Bug fixes
- The timestamp shown in the Updated field of the Deployment view in the Astro UI is now properly updated when you create or modify environment variables.
- Fixed an issue where logging in to the Airflow UI with unrecognized credentials could freeze you on an error page.
August 24, 2022
Additional improvements
- When you configure worker queues in the Astro UI, the total CPU and memory capacity of each worker instance type is now shown instead of the nominal available resources.
- Improved error handling for creating new worker queues when soft-deleted worker queues might still exist on the data plane.
Bug fixes
- Fixed an issue where running
astro deploy
with a Deployment API key could revert changes to a worker queue's size that were previously set in the Astro UI. - Fixed an issue where the Lineage tab in the Astro UI showed all job durations as having a length of 0.
August 18, 2022
Create multiple worker queues
Worker queues are a new way to configure your Deployment to best fit the needs of your tasks. A worker queue is a set of configurations that apply to a group of workers in your Deployment. Within a worker queue, you can configure worker type and size as well as autoscaling behavior. By configuring multiple worker queues for different types of tasks, you can better optimize for the performance, reliability, and throughput of your Deployment.
In the Astro UI, you can now create multiple worker queues. Once you create a worker queue, you can assign a task to that worker queue by adding a simple queue='<worker-queue-name>'
argument in your DAG code.
This feature enables the ability to:
- Use more than one worker type within a single Deployment and cluster. Previously, a single cluster on Astro supported only one worker type.
- Isolate long-running tasks from short-running tasks to avoid errors related to competing resource requests.
- Fine-tune autoscaling behavior for different groups of tasks within a single Deployment.
For example, if you have a task that requires significantly more CPU than memory, you can assign it to a queue that's configured with workers that are optimized for compute usage.
To learn more about configuring worker queues, see Configure Deployment resources.
New worker sizing
This Astro release introduces a new, simple way to allocate resources to the workers in your Deployment. Instead of choosing a varying combination of CPU and memory, you can now select a worker type in the Astro UI as long as it's enabled in your cluster. For example, m5.2xlarge
or c6i.8xlarge
on AWS. Once you select a worker type, Astronomer will create the biggest worker that that worker type can support to ensure that your tasks have enough resources to execute successfully.
Astro's worker sizing enables a few benefits:
- You can no longer configure a worker that is too large or otherwise not supported by your underlying cluster. Previously, misconfiguring worker size often resulted in task failures.
- A more efficient use of infrastructure. Astronomer has found that a lower number of larger workers is more efficient than a higher number of smaller workers.
- A higher level of reliability. This worker sizing model results in less volatility and a lower frequency of cluster autoscaling events, which lowers the frequency of errors such as zombie tasks and missing task logs.
- The legacy AU unit is no longer applicable in the context of the worker. You only have to think about CPU, memory, and worker type.
Worker sizing on Astro is now defined in the context of worker queues. For more information about worker sizing, see Configure Deployment resources. For a list of supported worker types, see the AWS, GCP, and Azure resource references.
New Maximum Tasks per Worker setting
A new Maximum Tasks per Worker configuration is now available in the Deployment view of the Astro UI. Maximum tasks per worker determines the maximum number of tasks that a single worker can process at a time and is the basis of worker autoscaling behavior. It is equivalent to worker concurrency in Apache Airflow.
Previously, maximum tasks per worker was permanently set to 16 and was not configurable on Astro. Now, you can set maximum tasks per worker anywhere between 1 and 64 based on the needs of your tasks. It can be set per worker queue on a Deployment.
To learn more, see Worker autoscaling logic.
New Worker Count (Min-Max) setting
A new Worker Count (Min-Max) configuration is now available in the Deployment view of the Astro UI. This value defines the minimum and maximum number of workers that can run at a time.
Use this setting to fine-tune worker autoscaling behavior in your Deployment. By default, the minimum number of workers is 1 and the maximum is 10.
Support for multiple Organizations
A single user account can now belong to multiple Organizations. A user with multiple Organizations can switch to another Organization by clicking on their current Organization's name in the Astro UI and then clicking Switch Organization.
Note that switching Organizations with the Astro CLI is not yet supported. For more information, see Switch Organizations.
New Azure region (Australia East)
You can now create an Astro cluster on Azure in Australia East (New South Wales).
New Google Cloud Platform regions
You can now create an Astro cluster on GCP in the following regions:
australia-southeast2
(Melbourne)asia-east1
(Taiwan)asia-south2
(Delhi)asia-southeast2
- (Jakarta)europe-north1
(Finland)europe-southwest1
(Madrid)europe-west8
(Milan)europe-west9
(Paris)northamerica-northeast2
(Toronto)southamerica-west1
(Santiago)us-east5
(Columbus)us-south1
(Dallas)
Bug fixes
- Fixed an issue where the Astro UI's Resource Settings page wasn't showing units for CPU and Memory values.
August 10, 2022
Updated user permissions for Organization and Workspace roles
The following user roles have new and modified permissions:
- Organization Owners now have Workspace Admin permissions for all Workspaces in their Organization. This role can now access Organization Workspaces, Deployments, and usage data.
- Organization Billing Admins can now view usage for all Workspaces in their Organization regardless of their Workspace permissions.
- Workspace Editors can now delete any Deployment in their Workspace.
Automatic access for new users authenticating with an identity provider
If your organization has implemented an identity provider (IdP), any new user who authenticates to Astro through your IdP is now automatically assigned the Organization Member role. This means that users authenticating through your IdP do not need to be invited by email before joining your Organization.
Additional improvements
- Added a security measure that ensures Workspace roles can only be assigned to users who have an Organization role in the Organization in which the Workspace is hosted. This ensures that a user who does not belong to your Organization cannot be assigned a Workspace role within it.
August 2, 2022
Support for Astro on Azure Kubernetes Service (AKS)
Astro now officially supports Astro clusters on AKS. This includes support for an initial set of AKS regions.
For more information about the installation process and supported configurations, see Install Astro on Azure and Resource Reference Azure.
Bug fixes
- Pending invites no longer appear for active users in the Astro UI.
July 27, 2022
New Deployment optimizations for high availability (HA)
This release introduces two changes that ensure a higher level of reliability for Deployments on Astro:
-
PgBouncer, a microservice that increases resilience by pooling database connections, is now considered highly available on Astro. Every Deployment must now have 2 PgBouncer Pods instead of 1, each assigned to a different node within the cluster. This change protects against pod-level connection issues resulting in zombie tasks, which was previously seen during cluster downscaling events. PgBouncer is fully managed by Astronomer and is not configurable.
-
The Airflow scheduler is now configured with an anti-affinity policy to limit the possibility of all schedulers for a single Deployment being impacted by an incident within a single node on an Astro cluster. For users who set Scheduler Count in the Astro UI to 2, this means that those 2 scheduler Pods cannot be assigned to the same node and instead require a minimum of 2 nodes total. To avoid significant increases in cost, 3 or 4 schedulers can share the same 2 nodes and will not necessarily result in a higher node count minimum.
For more information on Deployment configurations, see Deployment settings.
Additional improvements
- Added tooltips for Deployment overview metrics in the Astro UI.
July 21, 2022
Additional improvements
- You can now access an Organization's AWS external ID from the Settings tab of the Astro UI.
- Organizations now need only a single AWS external ID for all clusters. Previously, each cluster required a unique external ID, which added complexity to the installation and cluster creation process.
- You can now remove a user from an Organization from the Astro UI.
- Organization Billing Admins can now view task usage for all Workspaces regardless of their Workspace permissions.
July 14, 2022
Additional improvements
- The Astro UI Clusters page now includes the cluster ID value.
- Organization Owners and Organization Billing Admins can now update the Organization name in the Astro UI Settings page.
- The Astro UI Analytics page can now show data for the last 30 minutes.
Bug fixes
- When you change Worker Resources for a Deployment in the Astro UI, any errors related to your worker size request are now based on the correct node instance type that your cluster is running.
- When you select a Workspace and click Go back in a browser, the page now reloads as expected.
- A Page not found error message no longer appears when you select a Deployment in the Usage page of the Astro UI.
- The Deployment Analytics page now displays the correct date.
- Deprecated versions of Astro Runtime now appear correctly in the Deployment page of the Astro UI. Previously, versions were appended with
-deprecated
.
June 30, 2022
New Google Cloud Platform regions
You can now create an Astro cluster on GCP in the following regions:
asia-northeast1
(Tokyo)asia-northeast2
(Osaka)asia-northeast3
(Seoul)asia-south1
(Mumbai)europe-central2
(Warsaw)europe-west6
(Zurich)northamerica-northeast1
(Montreal)us-west3
(Salt Lake City)
Additional improvements
- You can now search for Organization members by name, email address, and role in the People tab of the Organization view in the Astro UI. You can also search for members in the Access tab of the Workspace view.
Bug fixes
- Fixed an issue where you could not use the KubernetesPodOperator to execute tasks in a Kubernetes cluster outside of your Astro cluster. See KubernetesPodOperator.
June 23, 2022
New Google Cloud Platform regions
You can now create an Astro cluster on GCP in the following regions:
asia-southeast1
(Singapore)australia-southeast1
(Sydney)europe-west1
(Belgium)europe-west2
(England)europe-west3
(Frankfurt)southamerica-east1
(São Paulo)us-west2
(Los Angeles)us-west4
(Nevada)
June 16, 2022
Submit Support Requests in the Astro UI
Support requests can now be created and submitted in the Astro UI. You no longer need to open an account on the Astronomer support portal to reach the Astronomer team. To streamline the request process, the Submit Support Request form auto-populates your currently selected Workspace and Deployment in the Astro UI.
Parallelism Now Autoscales with a Deployment's Worker Count
To better scale concurrent task runs, Astro now dynamically calculates parallelism
, which is an Airflow configuration that determines the maximum number of tasks that can run concurrently within a single Deployment.
A Deployment's parallelism
is now equal to the current number of workers multiplied by the worker_concurrency
value. This change ensures that your task runs won't be limited by a static parallelism limit as workers autoscale in your Deployment. See Worker Autoscaling Logic for more information.
Note that you can still use a static parallelism
value by setting AIRFLOW__CORE__PARALLELISM
as an environment variable.
Bug Fixes
- Fixed a rare issue where some user emails would be associated with the wrong username.
- Fixed an issue where you could not properly sort entries in the People tab by name.
June 9, 2022
Update Deployment configurations with the Astro CLI
You can now programmatically update the configurations for your Astro Deployments using Deployment API keys and the Astro CLI. Updating a Deployment with an API key doesn't require manual user authentication, meaning that you can now add Deployment configuration steps to automated processes such as CI/CD pipelines.
Specifically, you can now run the following commands with Deployment API keys:
astro deployment list
astro deployment update
astro deployment variable create
astro deployment variable list
astro deployment variable update
Bug fixes
- Fixed an issue where a Deployment's logs wouldn't load in the Astro UI if it was the only Deployment in the Workspace