actions/actions-runner-controller
1
0
Fork 0
mirror of https://github.com/actions/actions-runner-controller.git synced 2026-03-29 10:22:46 +08:00
Code Issues Packages Projects Releases Wiki Activity

Shutdown the scaleset when runner is deprecated (#4404)

Browse Source 
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
Nikola Jokic
2026-03-19 13:30:20 +01:00
committed by GitHub
parent 40595d806f
commit 9bc1c9e53e
18 changed files with 411 additions and 180 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
controllers/actions.github.com/ephemeralrunner_controller.go
View File

@@ -201,7 +201,7 @@ func (r *EphemeralRunnerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
}
}
if ephemeralRunner.Status.RunnerId == 0 {
if ephemeralRunner.Status.RunnerID == 0 {
log.Info("Updating ephemeral runner status with runnerId and runnerName")
runnerID, err := strconv.Atoi(string(secret.Data["runnerId"]))
if err != nil {
@@ -216,12 +216,12 @@ func (r *EphemeralRunnerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
runnerName := string(secret.Data["runnerName"])
if err := patchSubResource(ctx, r.Status(), ephemeralRunner, func(obj *v1alpha1.EphemeralRunner) {
obj.Status.RunnerId = runnerID
obj.Status.RunnerID = runnerID
obj.Status.RunnerName = runnerName
}); err != nil {
return ctrl.Result{}, fmt.Errorf("failed to update runner status for RunnerId/RunnerName/RunnerJITConfig: %w", err)
}
ephemeralRunner.Status.RunnerId = runnerID
ephemeralRunner.Status.RunnerID = runnerID
ephemeralRunner.Status.RunnerName = runnerName
log.Info("Updated ephemeral runner status with runnerId and runnerName")
}
@@ -324,7 +324,8 @@ func (r *EphemeralRunnerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
return ctrl.Result{}, r.deleteEphemeralRunnerOrPod(ctx, ephemeralRunner, pod, log)
}
if cs.State.Terminated.ExitCode == 0 {
switch cs.State.Terminated.ExitCode {
case 0:
log.Info("Runner container has succeeded but pod is in failed phase; Assume successful exit")
// If the pod is in a failed state, that means that at least one container exited with non-zero exit code.
// If the runner container exits with 0, we assume that the runner has finished successfully.
@@ -335,6 +336,12 @@ func (r *EphemeralRunnerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
return ctrl.Result{}, err
}
return ctrl.Result{}, nil
case 7:
if err := r.markAsOutdated(ctx, ephemeralRunner, log); err != nil {
log.Error(err, "Failed to set ephemeral runner to phase Outdated")
return ctrl.Result{}, err
}
return ctrl.Result{}, nil
}
log.Error(
@@ -357,6 +364,13 @@ func (r *EphemeralRunnerReconciler) Reconcile(ctx context.Context, req ctrl.Requ
}
return ctrl.Result{}, nil
case cs.State.Terminated.ExitCode == 7: // outdated
if err := r.markAsOutdated(ctx, ephemeralRunner, log); err != nil {
log.Error(err, "Failed to set ephemeral runner to phase Outdated")
return ctrl.Result{}, err
}
return ctrl.Result{}, nil
case cs.State.Terminated.ExitCode != 0: // failed
log.Info("Ephemeral runner container failed", "exitCode", cs.State.Terminated.ExitCode)
return ctrl.Result{}, r.deleteEphemeralRunnerOrPod(ctx, ephemeralRunner, pod, log)
@@ -390,7 +404,7 @@ func (r *EphemeralRunnerReconciler) deleteEphemeralRunnerOrPod(ctx context.Conte
log.Error(err, "Failed to get actions client for removing the runner from the service")
return nil
}
if err := actionsClient.RemoveRunner(ctx, int64(ephemeralRunner.Status.RunnerId)); err != nil {
if err := actionsClient.RemoveRunner(ctx, int64(ephemeralRunner.Status.RunnerID)); err != nil {
log.Error(err, "Failed to remove the runner from the service")
return nil
}
@@ -549,7 +563,7 @@ func (r *EphemeralRunnerReconciler) cleanupRunnerLinkedSecrets(ctx context.Conte
func (r *EphemeralRunnerReconciler) markAsFailed(ctx context.Context, ephemeralRunner *v1alpha1.EphemeralRunner, errMessage string, reason string, log logr.Logger) error {
log.Info("Updating ephemeral runner status to Failed")
if err := patchSubResource(ctx, r.Status(), ephemeralRunner, func(obj *v1alpha1.EphemeralRunner) {
obj.Status.Phase = corev1.PodFailed
obj.Status.Phase = v1alpha1.EphemeralRunnerPhaseFailed
obj.Status.Reason = reason
obj.Status.Message = errMessage
}); err != nil {
@@ -565,6 +579,24 @@ func (r *EphemeralRunnerReconciler) markAsFailed(ctx context.Context, ephemeralR
return nil
}
func (r *EphemeralRunnerReconciler) markAsOutdated(ctx context.Context, ephemeralRunner *v1alpha1.EphemeralRunner, log logr.Logger) error {
log.Info("Updating ephemeral runner status to Outdated")
if err := patchSubResource(ctx, r.Status(), ephemeralRunner, func(obj *v1alpha1.EphemeralRunner) {
obj.Status.Phase = v1alpha1.EphemeralRunnerPhaseOutdated
obj.Status.Reason = "Outdated"
obj.Status.Message = "Runner is deprecated"
}); err != nil {
return fmt.Errorf("failed to update ephemeral runner status Phase/Message: %w", err)
}
log.Info("Removing the runner from the service")
if err := r.deleteRunnerFromService(ctx, ephemeralRunner, log); err != nil {
return fmt.Errorf("failed to remove the runner from service: %w", err)
}
return nil
}
// deletePodAsFailed is responsible for deleting the pod and updating the .Status.Failures for tracking failure count.
// It should not be responsible for setting the status to Failed.
//
@@ -722,7 +754,7 @@ func (r *EphemeralRunnerReconciler) createPod(ctx context.Context, runner *v1alp
log.Info("Created ephemeral runner pod",
"runnerScaleSetId", runner.Spec.RunnerScaleSetID,
"runnerName", runner.Status.RunnerName,
"runnerId", runner.Status.RunnerId,
"runnerId", runner.Status.RunnerID,
"configUrl", runner.Spec.GitHubConfigUrl,
"podName", newPod.Name)
@@ -765,7 +797,8 @@ func (r *EphemeralRunnerReconciler) updateRunStatusFromPod(ctx context.Context,
}
}
phaseChanged := ephemeralRunner.Status.Phase != pod.Status.Phase
phase := v1alpha1.EphemeralRunnerPhase(pod.Status.Phase)
phaseChanged := ephemeralRunner.Status.Phase != phase
readyChanged := ready != ephemeralRunner.Status.Ready
if !phaseChanged && !readyChanged {
@@ -780,7 +813,7 @@ func (r *EphemeralRunnerReconciler) updateRunStatusFromPod(ctx context.Context,
"ready", ready,
)
err := patchSubResource(ctx, r.Status(), ephemeralRunner, func(obj *v1alpha1.EphemeralRunner) {
obj.Status.Phase = pod.Status.Phase
obj.Status.Phase = phase
obj.Status.Ready = ready
obj.Status.Reason = pod.Status.Reason
obj.Status.Message = pod.Status.Message
@@ -799,13 +832,13 @@ func (r *EphemeralRunnerReconciler) deleteRunnerFromService(ctx context.Context,
return fmt.Errorf("failed to get actions client for runner: %w", err)
}
log.Info("Removing runner from the service", "runnerId", ephemeralRunner.Status.RunnerId)
err = client.RemoveRunner(ctx, int64(ephemeralRunner.Status.RunnerId))
log.Info("Removing runner from the service", "runnerId", ephemeralRunner.Status.RunnerID)
err = client.RemoveRunner(ctx, int64(ephemeralRunner.Status.RunnerID))
if err != nil {
return fmt.Errorf("failed to remove runner from the service: %w", err)
}
log.Info("Removed runner from the service", "runnerId", ephemeralRunner.Status.RunnerId)
log.Info("Removed runner from the service", "runnerId", ephemeralRunner.Status.RunnerID)
return nil
}