Create backoff mechanism for failed runners and allow re-creation of failed ephemeral runners (#4059)

apis/actions.github.com/v1alpha1/ephemeralrunner_types.go

@@ -119,7 +119,7 @@ type EphemeralRunnerStatus struct {
 	RunnerJITConfig string `json:"runnerJITConfig,omitempty"`
 
 	// +optional
-	Failures map[string]bool `json:"failures,omitempty"`
+	Failures map[string]metav1.Time `json:"failures,omitempty"`
 
 	// +optional
 	JobRequestId int64 `json:"jobRequestId,omitempty"`
@@ -137,6 +137,20 @@ type EphemeralRunnerStatus struct {
 	JobDisplayName string `json:"jobDisplayName,omitempty"`
 }
 
+func (s *EphemeralRunnerStatus) LastFailure() metav1.Time {
+	var maxTime metav1.Time
+	if len(s.Failures) == 0 {
+		return maxTime
+	}
+
+	for _, ts := range s.Failures {
+		if ts.After(maxTime.Time) {
+			maxTime = ts
+		}
+	}
+	return maxTime
+}
+
 // +kubebuilder:object:root=true
 
 // EphemeralRunnerList contains a list of EphemeralRunner

apis/actions.github.com/v1alpha1/zz_generated.deepcopy.go

@@ -22,6 +22,7 @@ package v1alpha1
 
 import (
 	"k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	runtime "k8s.io/apimachinery/pkg/runtime"
 )
 
@@ -459,9 +460,9 @@ func (in *EphemeralRunnerStatus) DeepCopyInto(out *EphemeralRunnerStatus) {
 	*out = *in
 	if in.Failures != nil {
 		in, out := &in.Failures, &out.Failures
-		*out = make(map[string]bool, len(*in))
+		*out = make(map[string]metav1.Time, len(*in))
 		for key, val := range *in {
-			(*out)[key] = val
+			(*out)[key] = *val.DeepCopy()
 		}
 	}
 }