For AI agents: a documentation index is available at the root level at /llms.txt and /llms-full.txt. Append /llms.txt to any URL for a page-level index, or .md for the markdown version of any page.
Remote Execution uses a decoupled architecture where Astronomer manages the orchestration plane and the customer manages the execution plane. This document defines the responsibilities for each party.
Astronomer is responsible for managing the orchestration plane and supporting the Remote Execution platform, including:
Orchestration plane
Operating and maintaining the Airflow scheduler, API server, web server, and metadata database in Astro’s cloud infrastructure.
Managing the Remote Execution API that coordinates task distribution between the orchestration plane and customer-managed agents, and performs task lifecycle management.
Monitoring agent heartbeats and automatically rerouting tasks to healthy agents when an agent becomes unavailable.
Providing the Astro UI and Astro API for Deployment management, agent registration, and token creation.
Securing authentication and authorization for all orchestration plane interfaces, including the Astro UI, API, and CLI.
Maintaining data encryption at rest and in transit for all orchestration plane components.
Agent software and support
Publishing and maintaining Remote Execution Agent images.
Publishing and maintaining the Remote Execution Agent Helm chart.
Providing Astronomer support for Remote Execution configuration and troubleshooting.
Customer responsibilities
The customer is responsible for managing the execution plane and the infrastructure that Remote Execution Agents run on, including:
Kubernetes infrastructure
Provisioning and maintaining Kubernetes clusters where agents run.
Managing cluster capacity, node pools, and autoscaling to support agent workloads.
Applying Kubernetes and node OS security patches and version upgrades.
Configuring network policies, firewalls, and access controls within the cluster.
Ensuring network connectivity from agent clusters to the Astro orchestration plane. See Allowlist Astro domains.
Creating and managing agent tokens for authenticating agents to the orchestration plane.
Pulling agent images from the Astronomer registry and storing them in a private registry when required.
Building and maintaining custom agent images with additional Python packages and OS-level dependencies required by Dag code.
Upgrading agents to latest (recommended) or other supported versions within the maintenance window.
Monitoring agent Pod health and resource utilization in Kubernetes.
Secrets and credentials
Configuring a secrets backend (AWS Secrets Manager, Azure Key Vault, Google Cloud Secret Manager, or HashiCorp Vault) for Airflow connections and variables.
Managing and rotating credentials stored in the secrets backend.
Configuring workload identity or service account permissions for agents to access the secrets backend.
Securing network communications between agents and data resources in the execution plane, including secrets backends, object storage, and data sources.
Managing IP allowlists and firewall rules for agent clusters.
Implementing and maintaining secure, high-quality data pipelines, including dependency and vulnerability management.