feat: added config option to disable agent pod creation. Fixes argoproj#7891 (argoproj#15844)

Signed-off-by: workflow-automation <workflow-automation@msp.amadeus.net>
Co-authored-by: workflow-automation <workflow-automation@msp.amadeus.net>

Author: gaurang9991

.features/pending/disable-agent-pod.md

@@ -0,0 +1,6 @@
+Description: Disable agent pod creation for plugins
+Authors: [Gaurang Mishra](https://github.com/gaurang9991)
+Component: General
+Issues: 7891
+
+Allow users to disable agent pod creation for plugins. Workflow Controller watches the task sets updated by exeternal controllers or agents. User should be careful using this, when enabled it stop creating default agent pods for HTTP templates.

config/config.go

@@ -128,6 +128,11 @@ type Config struct {
 	// (e.g., due to Eviction, DiskPressure, Preemption). This allows recovery from transient
 	// infrastructure issues without requiring a retryStrategy on templates.
 	FailedPodRestart *FailedPodRestartConfig `json:"failedPodRestart,omitempty"`
+
+	// DisableAgentPodCreation disables the creation of agent pods for HTTP and Plugin templates.
+	// This is useful when external agents are responsible for executing these templates and the controller should not create agent pods.
+	// Note: when this is set to true, HTTP templates will not be reconciled and the controller will not attempt to create agent pods for them.
+	DisableAgentPodCreation bool `json:"disableAgentPodCreation,omitempty"`
 }
 
 // FailedPodRestartConfig configures automatic restart of pods that fail before entering Running state.

docs/workflow-controller-configmap.md

@@ -98,6 +98,7 @@ Config contains the root of the configuration settings for the workflow controll
 | `Synchronization`          | [`SyncConfig`](#syncconfig)                                                                                 | Synchronization via databases config                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        |
 | `ArtifactDrivers`          | `Array<`[`ArtifactDriver`](#artifactdriver)`>`                                                              | ArtifactDrivers lists artifact driver plugins we can use                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                    |
 | `FailedPodRestart`         | [`FailedPodRestartConfig`](#failedpodrestartconfig)                                                         | FailedPodRestart configures automatic restart of pods that fail before entering Running state (e.g., due to Eviction, DiskPressure, Preemption). This allows recovery from transient infrastructure issues without requiring a retryStrategy on templates.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |
+| `DisableAgentPodCreation`  | `bool`                                                                                                      | DisableAgentPodCreation disables the creation of agent pods for HTTP and Plugin templates. This is useful when external agents are responsible for executing these templates and the controller should not create agent pods. Note: when this is set to true, HTTP templates will not be reconciled and the controller will not attempt to create agent pods for them.                                                                                                                                                                                                                                                                                                                                                                                      |
 
 ## NodeEvents
 

docs/workflow-controller-configmap.yaml

@@ -486,3 +486,9 @@ data:
   #     Workflow cannot run an arbitrary Workflow, use this option.
   workflowRestrictions: |
     templateReferencing: Strict
+  
+  # disableAgentPodCreation disables the creation of agent pods for HTTP and Plugin templates. 
+  # This is useful when external agents are responsible for executing these templates and the controller should not create agent pods. 
+  # Note: when this is set to true, HTTP templates will not be reconciled and the controller will not attempt to create agent pods for them.
+  # Defaults to false
+  disableAgentPodCreation: "false"

workflow/controller/agent.go

@@ -110,6 +110,9 @@ func (woc *wfOperationCtx) getCertVolumeMount(ctx context.Context, name string)
 }
 
 func (woc *wfOperationCtx) createAgentPod(ctx context.Context) (*apiv1.Pod, error) {
+	if woc.controller.Config.DisableAgentPodCreation {
+		return nil, nil
+	}
 	podName := woc.getAgentPodName()
 	ctx, log := woc.log.WithField("podName", podName).InContext(ctx)
 

workflow/controller/agent_test.go

@@ -171,3 +171,46 @@ func TestAssessAgentPodStatus(t *testing.T) {
 		assert.Empty(t, msg)
 	})
 }
+
+func TestDisableAgentPodCreation(t *testing.T) {
+	ctx := logging.TestContext(t.Context())
+	wf := wfv1.MustUnmarshalWorkflow(`apiVersion: argoproj.io/v1alpha1
+kind: Workflow
+metadata:
+  name: http-template
+  namespace: default
+spec:
+  podSpecPatch: |
+    nodeName: virtual-node
+  entrypoint: main
+  templates:
+    - name: main
+      steps:
+        - - name: good
+            template: http
+            arguments:
+              parameters: [{name: url, value: "https://raw.githubusercontent.com/argoproj/argo-workflows/4e450e250168e6b4d51a126b784e90b11a0162bc/pkg/apis/workflow/v1alpha1/generated.swagger.json"}]
+        - - name: bad
+            template: http
+            continueOn:
+              failed: true
+            arguments:
+              parameters: [{name: url, value: "http://openlibrary.org/people/george08/nofound.json"}]
+
+    - name: http
+      inputs:
+        parameters:
+          - name: url
+      http:
+       url: "{{inputs.parameters.url}}"
+
+`)
+	cancel, controller := newController(ctx, wf, defaultServiceAccount)
+	woc := newWorkflowOperationCtx(ctx, wf, controller)
+	woc.controller.Config.DisableAgentPodCreation = true
+	defer cancel()
+	woc.operate(ctx)
+	pods, err := woc.controller.kubeclientset.CoreV1().Pods("default").List(ctx, v1.ListOptions{})
+	require.NoError(t, err)
+	assert.Empty(t, pods.Items)
+}