feat: Add DagProcessor and Triggerer roles

CHANGELOG.md

@@ -5,6 +5,7 @@
 ### Added
 
 - Add a flag to determine if database initialization steps should be executed ([#669]).
+- Add new roles for dag-processor and triggerer processes ([#679]).
 
 ### Fixed
 
@@ -19,6 +20,7 @@
 [#668]: https://github.com/stackabletech/airflow-operator/pull/668
 [#669]: https://github.com/stackabletech/airflow-operator/pull/669
 [#678]: https://github.com/stackabletech/airflow-operator/pull/678
+[#679]: https://github.com/stackabletech/airflow-operator/pull/679
 [#683]: https://github.com/stackabletech/airflow-operator/pull/683
 
 ## [25.7.0] - 2025-07-23

docs/modules/airflow/examples/getting_started/code/airflow.yaml

@@ -19,8 +19,16 @@ spec:
   celeryExecutors:
     roleGroups:
       default:
-        replicas: 2
+        replicas: 1
   schedulers:
     roleGroups:
       default:
         replicas: 1
+  dagProcessors:
+    roleGroups:
+      default:
+        replicas: 1
+  triggerers:
+    roleGroups:
+      default:
+        replicas: 1

docs/modules/airflow/examples/getting_started/code/getting_started.sh

@@ -88,6 +88,8 @@ echo "Awaiting Airflow rollout finish ..."
 kubectl rollout status --watch --timeout=5m statefulset/airflow-webserver-default
 kubectl rollout status --watch --timeout=5m statefulset/airflow-worker-default
 kubectl rollout status --watch --timeout=5m statefulset/airflow-scheduler-default
+kubectl rollout status --watch --timeout=5m statefulset/airflow-dagprocessor-default
+kubectl rollout status --watch --timeout=5m statefulset/airflow-triggerer-default
 # end::watch-airflow-rollout[]
 
 echo "Starting port-forwarding of port 8080"

docs/modules/airflow/examples/getting_started/code/getting_started.sh.j2

@@ -88,6 +88,8 @@ echo "Awaiting Airflow rollout finish ..."
 kubectl rollout status --watch --timeout=5m statefulset/airflow-webserver-default
 kubectl rollout status --watch --timeout=5m statefulset/airflow-worker-default
 kubectl rollout status --watch --timeout=5m statefulset/airflow-scheduler-default
+kubectl rollout status --watch --timeout=5m statefulset/airflow-dagprocessor-default
+kubectl rollout status --watch --timeout=5m statefulset/airflow-triggerer-default
 # end::watch-airflow-rollout[]
 
 echo "Starting port-forwarding of port 8080"

docs/modules/airflow/pages/getting_started/first_steps.adoc

@@ -38,11 +38,15 @@ NOTE: The admin user is disabled if you use a non-default authentication mechani
 
 == Airflow
 
-An Airflow cluster is made of up three components:
+An Airflow cluster is made up of several components, two of which are optional:
 
 * `webserver`: this provides the main UI for user-interaction
 * `executors`: the CeleryExecutor or KubernetesExecutor nodes over which the job workload is distributed by the scheduler
 * `scheduler`: responsible for triggering jobs and persisting their metadata to the backend database
+* `dagProcessors`: (Optional) responsible for monitoring, parsing and preparing DAGs for processing.
+If this role is not specified then the process will be started as a scheduler subprocess (Airflow 2.x), or as a standalone process in the same container as the scheduler (Airflow 3.x+)
+* `triggerers`: (Optional) DAGs making use of deferrable operators can be used together with one or more triggerer processes to free up worker slots.
+This deferral process is also useful for providing a measure of high availability
 
 Create a file named `airflow.yaml` with the following contents:
 
@@ -92,7 +96,9 @@ airflow-redis-master            1/1     16m
 airflow-redis-replicas          1/1     16m
 airflow-scheduler-default       1/1     11m
 airflow-webserver-default       1/1     11m
-airflow-celery-executor-default 2/2     11m
+airflow-celery-executor-default 1/1     11m
+airflow-dagprocessor-default    1/1     11m
+airflow-triggerer-default       1/1     11m
 ----
 
 When the Airflow cluster has been created and the database is initialized, Airflow can be opened in the

docs/modules/airflow/pages/index.adoc

@@ -3,11 +3,11 @@
 :keywords: Stackable Operator, Apache Airflow, Kubernetes, k8s, operator, job pipeline, scheduler, ETL
 :airflow: https://airflow.apache.org/
 :dags: https://airflow.apache.org/docs/apache-airflow/stable/core-concepts/dags.html
-:k8s-crs: https://kubernetes.io/docs/concepts/extend-kubernetes/api-extension/custom-resources/
 :github: https://github.com/stackabletech/airflow-operator/
 :crd: {crd-docs-base-url}/airflow-operator/{crd-docs-version}/
 :crd-airflowcluster: {crd-docs}/airflow.stackable.tech/airflowcluster/v1alpha1/
 :feature-tracker: https://features.stackable.tech/unified
+:deferrable-operators: https://airflow.apache.org/docs/apache-airflow/stable/authoring-and-scheduling/deferring.html#deferrable-operators-triggers
 
 [.link-bar]
 * {github}[GitHub {external-link-icon}^]
@@ -27,7 +27,8 @@ It guides you through installing the operator alongside a PostgreSQL database an
 === Custom resources
 
 The AirflowCluster is the resource for the configuration of the Airflow instance.
-The resource defines three xref:concepts:roles-and-role-groups.adoc[roles]: `webserver`, `worker` and `scheduler` (the `worker` role is embedded within `spec.celeryExecutors`: this is described in the next section).
+The custom resource defines the following xref:concepts:roles-and-role-groups.adoc[roles]: `webserver`, `worker`, `scheduler`, `dagProcessor` and `triggerer` (the `worker` role is embedded within `spec.celeryExecutors`: this is described in the next section).
+The `dagProcessor` and `triggerer` roles are optional.
 The various configuration options are explained in the xref:usage-guide/index.adoc[].
 It helps you tune your cluster to your needs by configuring xref:usage-guide/storage-resources.adoc[resource usage], xref:usage-guide/security.adoc[security], xref:usage-guide/logging.adoc[logging] and more.
 
@@ -70,6 +71,17 @@ kubernetesExecutors:
       ...
 ----
 
+=== DAG-Processors
+
+In Airflow 2.x, a DAG-Processor can be started either as a standalone process or a subprocess within the scheduler component.
+For Airflow 3.x+ it _must_ be started as a standalone process, either in a separate container or in the scheduler container.
+In each case the default will be applied (subprocess or combined in the scheduler container) if the role is not specified.
+
+=== Triggerers
+
+DAGs using deferrable operators can be combined with the triggerer component to free up worker slots and/or provide high availability.
+For more information, please refer to the {deferrable-operators}[documentation {external-link-icon}^].
+
 === Kubernetes resources
 
 Based on the custom resources you define, the operator creates ConfigMaps, StatefulSets and Services.

docs/modules/airflow/pages/usage-guide/storage-resources.adoc

@@ -1,14 +1,15 @@
 = Resource Requests
-:description: Find out about minimal HA Airflow requirements for CPU and memory, with defaults for schedulers, Celery executors, webservers using Kubernetes resource limits.
+:description: Find out about minimal HA Airflow requirements for CPU and memory, with defaults for schedulers, Celery executors, webservers, dagProcessors and triggerers using Kubernetes resource limits.
 
 include::home:concepts:stackable_resource_requests.adoc[]
 
-A minimal HA setup consisting of 2 schedulers, 2 workers and 2 webservers has the following https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/[resource requirements]:
+A minimal HA setup consisting of 2 schedulers, 2 workers, 2 webservers, 2 dag-processors and 1 triggerer has the following https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/[resource requirements]:
 
-* `8700m` CPU request
-* `17400m` CPU limit
-* `15872Mi` memory request and limit
+* `11600` CPU request
+* `23200` CPU limit
+* `18432Mi` memory request and limit
 
+This includes auxiliary containers for logging, metrics, and gitsync.
 Corresponding to the values above, the operator uses the following resource defaults:
 
 [source,yaml]
@@ -22,6 +23,9 @@ spec:
           max: "2"
         memory:
           limit: 1Gi
+    roleGroups:
+      default:
+        replicas: 2
   celeryExecutors:
     config:
       resources:
@@ -30,6 +34,9 @@ spec:
           max: "2"
         memory:
           limit: 3Gi
+    roleGroups:
+      default:
+        replicas: 2
   webservers:
     config:
       resources:
@@ -38,4 +45,29 @@ spec:
           max: "2"
         memory:
           limit: 3Gi
+    roleGroups:
+      default:
+        replicas: 2
+  dagProcessors:
+    config:
+      resources:
+        cpu:
+          min: "1"
+          max: "2"
+        memory:
+          limit: 1Gi
+    roleGroups:
+      default:
+        replicas: 2
+  triggerers:
+    config:
+      resources:
+        cpu:
+          min: "1"
+          max: "2"
+        memory:
+          limit: 1Gi
+    roleGroups:
+      default:
+        replicas: 1
 ----

rust/operator-binary/src/airflow_controller.rs

@@ -1156,7 +1156,10 @@ fn build_server_rolegroup_statefulset(
                 AirflowRole::Scheduler => {
                     "OrderedReady" // Scheduler pods should start after another, since part of their startup phase is initializing the database, see crd/src/lib.rs
                 }
-                AirflowRole::Webserver | AirflowRole::Worker => "Parallel",
+                AirflowRole::Webserver
+                | AirflowRole::Worker
+                | AirflowRole::DagProcessor
+                | AirflowRole::Triggerer => "Parallel",
             }
             .to_string(),
         ),