🌱 OPRUN-4795: Cover COD progress-deadline lifecycle

test/e2e/object-controller/features/cod_progress_deadline.feature

@@ -0,0 +1,184 @@
+@ObjectController @ClusterObjectDeployment @ProgressDeadline @CODDeadlineLifecycle
+Feature: ClusterObjectDeployment progress deadline lifecycle
+
+  A revision must recover within its deadline without creating another revision.
+  After completion, health may regress while completion remains recorded.
+  The deployment must keep reporting a successful rollout beyond the deadline.
+
+  Background:
+    Given object-controller is available
+    And namespace "${TEST_NAMESPACE}" is available
+
+  Scenario: Completed revision stays progressed when health regresses beyond the deadline
+    When resource is applied
+      """
+      apiVersion: olm.operatorframework.io/v1
+      kind: ClusterObjectDeployment
+      metadata:
+        name: ${NAME}
+      spec:
+        progressDeadlineMinutes: 1
+        template:
+          spec:
+            collisionProtection: Prevent
+            progressionProbes:
+            - selector:
+                type: Label
+                label:
+                  matchLabels:
+                    probe: deadline-gate
+              assertions:
+              - type: FieldValue
+                fieldValue:
+                  fieldPath: data.ready
+                  value: "true"
+            phases:
+            - name: resources
+              objects:
+              - type: InlineManifest
+                inlineManifest:
+                  apiVersion: v1
+                  kind: ConfigMap
+                  metadata:
+                    name: deadline-gate
+                    namespace: ${TEST_NAMESPACE}
+                    labels:
+                      probe: deadline-gate
+                  data:
+                    payload: unchanged
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as False with Reason ObjectsNotReady
+    And resource "clusterobjectset/${NAME}-1" matches
+      """
+      status:
+        completedAt: null
+      """
+    # Only the test owns data.ready, which is absent from the immutable manifest.
+    When resource is applied
+      """
+      apiVersion: v1
+      kind: ConfigMap
+      metadata:
+        name: deadline-gate
+        namespace: ${TEST_NAMESPACE}
+      data:
+        ready: "true"
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as True with Reason AllObjectsReady
+    And ClusterObjectDeployment "${NAME}" reports Progressing as False with Reason NewClusterObjectSetProgressed
+    And ClusterObjectDeployment "${NAME}" reports Available as True with Reason Available
+    And ClusterObjectDeployment "${NAME}" owns 1 ClusterObjectSets
+    And ClusterObjectSet "${NAME}-1" has a completion timestamp
+    When resource is applied
+      """
+      apiVersion: v1
+      kind: ConfigMap
+      metadata:
+        name: deadline-gate
+        namespace: ${TEST_NAMESPACE}
+      data:
+        ready: "false"
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as False with Reason ObjectsNotReady
+    And ClusterObjectDeployment "${NAME}" reports Available as False with Reason Unavailable
+    And ClusterObjectSet "${NAME}-1" has a completion timestamp
+    # Both controllers have reflected the health regression. The successful
+    # rollout reason must persist beyond the one-minute deadline.
+    And ClusterObjectDeployment "${NAME}" consistently reports Progressing as False with Reason NewClusterObjectSetProgressed for "70s"
+    When resource is applied
+      """
+      apiVersion: v1
+      kind: ConfigMap
+      metadata:
+        name: deadline-gate
+        namespace: ${TEST_NAMESPACE}
+      data:
+        ready: "true"
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as True with Reason AllObjectsReady
+    And ClusterObjectDeployment "${NAME}" reports Available as True with Reason Available
+    And ClusterObjectDeployment "${NAME}" reports Progressing as False with Reason NewClusterObjectSetProgressed
+    And ClusterObjectSet "${NAME}-1" has a completion timestamp
+    And ClusterObjectDeployment "${NAME}" owns 1 ClusterObjectSets
+    And resource "configmap/deadline-gate" matches
+      """
+      data:
+        payload: unchanged
+        ready: "true"
+      """
+
+  Scenario: Revision recovers before its original progress deadline
+    When resource is applied
+      """
+      apiVersion: olm.operatorframework.io/v1
+      kind: ClusterObjectDeployment
+      metadata:
+        name: ${NAME}
+      spec:
+        progressDeadlineMinutes: 1
+        template:
+          spec:
+            collisionProtection: Prevent
+            progressionProbes:
+            - selector:
+                type: Label
+                label:
+                  matchLabels:
+                    probe: deadline-gate
+              assertions:
+              - type: FieldValue
+                fieldValue:
+                  fieldPath: data.ready
+                  value: "true"
+            phases:
+            - name: resources
+              objects:
+              - type: InlineManifest
+                inlineManifest:
+                  apiVersion: v1
+                  kind: ConfigMap
+                  metadata:
+                    name: deadline-gate
+                    namespace: ${TEST_NAMESPACE}
+                    labels:
+                      probe: deadline-gate
+                  data:
+                    payload: unchanged
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as False with Reason ObjectsNotReady
+    And ClusterObjectDeployment "${NAME}" reports Progressing as True with Reason NewClusterObjectSetProgressing
+    And ClusterObjectDeployment "${NAME}" reports Available as False with Reason Unavailable
+    And ClusterObjectDeployment "${NAME}" owns 1 ClusterObjectSets
+    And resource "clusterobjectset/${NAME}-1" matches
+      """
+      status:
+        completedAt: null
+      """
+    When resource is applied
+      """
+      apiVersion: v1
+      kind: ConfigMap
+      metadata:
+        name: deadline-gate
+        namespace: ${TEST_NAMESPACE}
+      data:
+        ready: "true"
+      """
+    Then ClusterObjectSet "${NAME}-1" reports Ready as True with Reason AllObjectsReady
+    And ClusterObjectDeployment "${NAME}" reports Progressing as False with Reason NewClusterObjectSetProgressed
+    And ClusterObjectDeployment "${NAME}" reports Available as True with Reason Available
+    And ClusterObjectDeployment "${NAME}" owns 1 ClusterObjectSets
+    And resource "clusterobjectdeployment/${NAME}" matches
+      """
+      spec:
+        progressDeadlineMinutes: 1
+      """
+    # Compare the named revision's completion against its server-assigned creation
+    # timestamp plus the unchanged one-minute policy.
+    And ClusterObjectSet "${NAME}-1" completed before the deadline of ClusterObjectDeployment "${NAME}"
+    And resource "configmap/deadline-gate" matches
+      """
+      data:
+        payload: unchanged
+        ready: "true"
+      """

test/internal/e2e/object-controller/steps/deadline_steps.go

@@ -0,0 +1,69 @@
+package steps
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"time"
+
+	ocv1 "github.com/operator-framework/operator-controller/api/v1"
+	shared "github.com/operator-framework/operator-controller/test/internal/e2e/steps"
+)
+
+func readDeadlineObject(ctx context.Context, kind, name string, into any) error {
+	name = shared.SubstituteScenarioVars(name, shared.Scenario(ctx))
+	out, err := shared.K8sClient(ctx, "get", kind, name, "-o", "json")
+	if err != nil {
+		return fmt.Errorf("reading %s %q: %w", kind, name, err)
+	}
+	if err = json.Unmarshal([]byte(out), into); err != nil {
+		return fmt.Errorf("decoding %s %q: %w", kind, name, err)
+	}
+	return nil
+}
+
+// ClusterObjectSetHasCompletion checks presence only. API validation separately
+// guarantees that completedAt cannot be changed or removed once set.
+func ClusterObjectSetHasCompletion(ctx context.Context, name string) error {
+	var cos ocv1.ClusterObjectSet
+	if err := readDeadlineObject(ctx, "clusterobjectset", name, &cos); err != nil {
+		return err
+	}
+	if cos.Status.CompletedAt.IsZero() {
+		return fmt.Errorf("revision %q has no completedAt", cos.Name)
+	}
+	return nil
+}
+
+// ClusterObjectSetCompletedBeforeDeadline compares durable timestamps on the
+// named revision. The feature separately checks sole ownership.
+func ClusterObjectSetCompletedBeforeDeadline(ctx context.Context, name, deployment string) error {
+	var cos ocv1.ClusterObjectSet
+	if err := readDeadlineObject(ctx, "clusterobjectset", name, &cos); err != nil {
+		return err
+	}
+	var cod ocv1.ClusterObjectDeployment
+	if err := readDeadlineObject(ctx, "clusterobjectdeployment", deployment, &cod); err != nil {
+		return err
+	}
+	if cos.Spec.Group != cod.Name {
+		return fmt.Errorf("revision %q is not in deployment group %q", cos.Name, cod.Name)
+	}
+	if cos.CreationTimestamp.IsZero() {
+		return fmt.Errorf("revision %q has no creation timestamp", cos.Name)
+	}
+	if cos.Status.CompletedAt.IsZero() {
+		return fmt.Errorf("revision %q has no completedAt", cos.Name)
+	}
+	if cos.Status.CompletedAt.Before(&cos.CreationTimestamp) {
+		return fmt.Errorf("completedAt precedes revision creation")
+	}
+	if cod.Spec.ProgressDeadlineMinutes <= 0 {
+		return fmt.Errorf("deployment %q has no enabled deadline", cod.Name)
+	}
+	deadline := cos.CreationTimestamp.Add(time.Duration(cod.Spec.ProgressDeadlineMinutes) * time.Minute)
+	if !cos.Status.CompletedAt.Time.Before(deadline) {
+		return fmt.Errorf("revision completedAt %s must be before deadline %s", cos.Status.CompletedAt, deadline)
+	}
+	return nil
+}

test/internal/e2e/object-controller/steps/deadline_steps_test.go

@@ -0,0 +1,123 @@
+package steps
+
+import (
+	"context"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/cucumber/godog"
+	"github.com/spf13/pflag"
+	"github.com/stretchr/testify/require"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+	ocv1 "github.com/operator-framework/operator-controller/api/v1"
+	shared "github.com/operator-framework/operator-controller/test/internal/e2e/steps"
+)
+
+// Only the API transport is replaced. Completion checks use the real helpers;
+// ordinary count/condition steps retain their existing tests.
+func deadlineFixture(t *testing.T) (context.Context, *ocv1.ClusterObjectDeployment, *ocv1.ClusterObjectSet, string) {
+	t.Helper()
+	ctx, err := shared.CreateScenarioContext(t.Context(), &godog.Scenario{Id: "1", Uri: "cod_progress_deadline.feature"})
+	require.NoError(t, err)
+	sc := shared.Scenario(ctx)
+	cod := &ocv1.ClusterObjectDeployment{ObjectMeta: metav1.ObjectMeta{Name: sc.ClusterExtensionName}}
+	cod.Spec.ProgressDeadlineMinutes = 1
+	cos := &ocv1.ClusterObjectSet{ObjectMeta: metav1.ObjectMeta{Name: cod.Name + "-1", CreationTimestamp: metav1.NewTime(time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC))}}
+	cos.Spec.Group = cod.Name
+	cos.Status.CompletedAt = metav1.NewTime(cos.CreationTimestamp.Add(10 * time.Second))
+	dir := t.TempDir()
+	t.Setenv("DEADLINE_FIXTURES", dir)
+	script := `#!/bin/sh
+if [ -f "$DEADLINE_FIXTURES/api-error" ]; then echo "API unavailable" >&2; exit 1; fi
+case "$1 $2" in
+ "get clusterobjectdeployment") cat "$DEADLINE_FIXTURES/cod.json";;
+ "get clusterobjectset") cat "$DEADLINE_FIXTURES/cos.json";;
+ *) echo "unexpected command: $*" >&2; exit 1;;
+esac
+`
+	cli := filepath.Join(dir, "kubectl")
+	//nolint:gosec // Private executable fixture.
+	require.NoError(t, os.WriteFile(cli, []byte(script), 0o700))
+	oldCLI := pflag.Lookup("k8s.cli").Value.String()
+	require.NoError(t, pflag.Set("k8s.cli", cli))
+	t.Cleanup(func() { require.NoError(t, pflag.Set("k8s.cli", oldCLI)) })
+	writeDeadlineFixture(t, dir, cod, cos)
+	return ctx, cod, cos, dir
+}
+
+func writeDeadlineJSON(t *testing.T, dir, name string, obj any) {
+	t.Helper()
+	data, err := json.Marshal(obj)
+	require.NoError(t, err)
+	require.NoError(t, os.WriteFile(filepath.Join(dir, name), data, 0o600))
+}
+func writeDeadlineFixture(t *testing.T, dir string, cod *ocv1.ClusterObjectDeployment, cos *ocv1.ClusterObjectSet) {
+	t.Helper()
+	writeDeadlineJSON(t, dir, "cod.json", cod)
+	writeDeadlineJSON(t, dir, "cos.json", cos)
+}
+
+func TestDeadlineTimestampComparison(t *testing.T) {
+	for _, tc := range []struct {
+		name   string
+		mutate func(*ocv1.ClusterObjectDeployment, *ocv1.ClusterObjectSet)
+		want   string
+	}{
+		{name: "before expiry"},
+		{name: "exact expiry", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) {
+			c.Status.CompletedAt = metav1.NewTime(c.CreationTimestamp.Add(time.Minute))
+		}, want: "before"},
+		{name: "after expiry", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) {
+			c.Status.CompletedAt = metav1.NewTime(c.CreationTimestamp.Add(61 * time.Second))
+		}, want: "before"},
+		{name: "before creation", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) {
+			c.Status.CompletedAt = metav1.NewTime(c.CreationTimestamp.Add(-time.Second))
+		}, want: "creation"},
+		{name: "missing creation", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) { c.CreationTimestamp = metav1.Time{} }, want: "creation"},
+		{name: "missing completion", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) { c.Status.CompletedAt = metav1.Time{} }, want: "completedAt"},
+		{name: "disabled deadline", mutate: func(c *ocv1.ClusterObjectDeployment, _ *ocv1.ClusterObjectSet) { c.Spec.ProgressDeadlineMinutes = 0 }, want: "deadline"},
+		{name: "different group", mutate: func(_ *ocv1.ClusterObjectDeployment, c *ocv1.ClusterObjectSet) { c.Spec.Group = "another" }, want: "group"},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			ctx, cod, cos, dir := deadlineFixture(t)
+			if tc.mutate != nil {
+				tc.mutate(cod, cos)
+			}
+			writeDeadlineFixture(t, dir, cod, cos)
+			err := ClusterObjectSetCompletedBeforeDeadline(ctx, "${NAME}-1", "${NAME}")
+			if tc.want == "" {
+				require.NoError(t, err)
+			} else {
+				require.ErrorContains(t, err, tc.want)
+			}
+		})
+	}
+}
+
+func TestClusterObjectSetHasCompletion(t *testing.T) {
+	for _, mode := range []string{"present", "missing", "API error", "invalid JSON"} {
+		t.Run(mode, func(t *testing.T) {
+			ctx, cod, cos, dir := deadlineFixture(t)
+			if mode == "missing" {
+				cos.Status.CompletedAt = metav1.Time{}
+			}
+			writeDeadlineFixture(t, dir, cod, cos)
+			if mode == "API error" {
+				require.NoError(t, os.WriteFile(filepath.Join(dir, "api-error"), nil, 0o600))
+			}
+			if mode == "invalid JSON" {
+				require.NoError(t, os.WriteFile(filepath.Join(dir, "cos.json"), []byte("invalid"), 0o600))
+			}
+			err := ClusterObjectSetHasCompletion(ctx, "${NAME}-1")
+			if mode == "present" {
+				require.NoError(t, err)
+			} else {
+				require.Error(t, err)
+			}
+		})
+	}
+}

test/internal/e2e/object-controller/steps/steps.go

@@ -22,6 +22,8 @@ import (
 // RegisterSteps registers ClusterObjectSet and ClusterObjectDeployment assertions.
 func RegisterSteps(sc *godog.ScenarioContext) {
 	sc.Step(`^object-controller is available$`, ObjectControllerIsAvailable)
+	sc.Step(`^ClusterObjectSet "([^"]+)" has a completion timestamp$`, ClusterObjectSetHasCompletion)
+	sc.Step(`^ClusterObjectSet "([^"]+)" completed before the deadline of ClusterObjectDeployment "([^"]+)"$`, ClusterObjectSetCompletedBeforeDeadline)
 	sc.Step(`^(?i)ClusterObjectSet "([^"]+)" lifecycle is set to "([^"]+)"$`, ClusterObjectSetLifecycleUpdate)
 	sc.Step(`^(?i)ClusterObjectSet "([^"]+)" reports ([[:alnum:]]+) as ([[:alnum:]]+) with Reason ([[:alnum:]]+)$`, ClusterObjectSetReportsConditionWithoutMsg)
 	sc.Step(`^(?i)ClusterObjectSet "([^"]+)" reports ([[:alnum:]]+) as ([[:alnum:]]+) with Reason ([[:alnum:]]+) and Message:$`, ClusterObjectSetReportsConditionWithMsg)