fix(operator): wake up for idle auto-stop — the timer had no driver
EmptySince was stamped and then never revisited: player joins/leaves do not touch the CRD, RCON is only probed inside Reconcile, and a steady Running server produces no watch events (its status update goes out unchanged and is a no-op). Live, an empty server with a 30s grace sat Running for minutes with zero reconciles in the log — the auto-stop existed only on paper. reconcileRunning now returns a RequeueAfter for idle-enabled servers: exactly at the deadline while empty, or a 30s probe cadence while occupied so the moment the last player leaves is noticed. New unit tests pin all three: deadline requeue, occupied cadence, and no requeue when disabled.
This commit is contained in:
2 files changed
+80
-1
No files matched your search
@@ -35,6 +35,7 @@ const (
|
||||
requeueStarting = 2 * time.Second
|
||||
requeueStopping = 5 * time.Second
|
||||
requeueSecret = 10 * time.Second
|
||||
requeueIdleProbe = 30 * time.Second
|
||||
defaultTimeoutSeconds = 300
|
||||
defaultReadinessTimeoutSec = 300
|
||||
)
|
||||
@@ -197,7 +198,25 @@ func (r *Reconciler) reconcileRunning(ctx context.Context, server *v1alpha1.Mine
|
||||
}
|
||||
|
||||
r.markRunningReady(server, players, endpointAddress)
|
||||
return ctrl.Result{}, r.patchStatus(ctx, server)
|
||||
if err := r.patchStatus(ctx, server); err != nil {
|
||||
return ctrl.Result{}, err
|
||||
}
|
||||
// Idle auto-stop has no natural wake-up: player joins/leaves never touch
|
||||
// this CRD and RCON is only probed here, so without a requeue the
|
||||
// empty-duration counter would be stamped once and then never revisited
|
||||
// (observed live: the stamp sat unexamined for minutes). Wake at the exact
|
||||
// deadline while the tally says empty, or on a slow cadence while players
|
||||
// are online, to notice the moment the last one leaves.
|
||||
if server.Spec.Rcon.Enabled && server.Spec.Idle.AutoStopEnabled && server.Spec.Idle.EmptySecondsBeforeStop > 0 {
|
||||
if server.Status.EmptySince != nil {
|
||||
deadline := server.Status.EmptySince.Time.Add(time.Duration(server.Spec.Idle.EmptySecondsBeforeStop) * time.Second)
|
||||
if wait := deadline.Sub(r.now().Time); wait > 0 {
|
||||
return ctrl.Result{RequeueAfter: wait}, nil
|
||||
}
|
||||
}
|
||||
return ctrl.Result{RequeueAfter: requeueIdleProbe}, nil
|
||||
}
|
||||
return ctrl.Result{}, nil
|
||||
}
|
||||
|
||||
func (r *Reconciler) reconcileStopped(ctx context.Context, server *v1alpha1.MinecraftServer) (ctrl.Result, error) {
|
||||
|
||||
@@ -427,6 +427,66 @@ func TestIdleAutoStop_ResetsWhenPlayerJoins(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// TestIdleAutoStop_RequeuesUntilDeadline pins the self-driving requeue: an
|
||||
// empty Running server must wake the controller at the auto-stop deadline with
|
||||
// no external event to lean on. Live, the EmptySince stamp sat unexamined for
|
||||
// minutes because nothing re-triggered the reconcile loop — this test fails if
|
||||
// the requeue is ever dropped again.
|
||||
func TestIdleAutoStop_RequeuesUntilDeadline(t *testing.T) {
|
||||
r, c := newReconciler(t, fakeProber{players: operator.PlayerCount{Online: 0, Max: 20}}, runningServer(), rconSecret())
|
||||
base := time.Date(2026, 7, 1, 10, 0, 0, 0, time.UTC)
|
||||
r.Now = func() metav1.Time { return metav1.NewTime(base) }
|
||||
|
||||
s := getServer(t, c, "survival")
|
||||
s.Spec.Idle = v1alpha1.IdleSpec{AutoStopEnabled: true, EmptySecondsBeforeStop: 30}
|
||||
if err := c.Update(context.Background(), s); err != nil {
|
||||
t.Fatalf("enable idle: %v", err)
|
||||
}
|
||||
|
||||
reconcile(t, r, "survival")
|
||||
markPodReady(t, c, "survival")
|
||||
res := reconcile(t, r, "survival")
|
||||
|
||||
if res.RequeueAfter != 30*time.Second {
|
||||
t.Fatalf("RequeueAfter = %v, want exactly 30s (wake at the auto-stop deadline)", res.RequeueAfter)
|
||||
}
|
||||
}
|
||||
|
||||
// TestIdleAutoStop_RequeuesWhileOccupied verifies the slow probe cadence that
|
||||
// notices the last player leaving: with players online there is no deadline to
|
||||
// aim at, but the tally must still be re-sampled.
|
||||
func TestIdleAutoStop_RequeuesWhileOccupied(t *testing.T) {
|
||||
r, c := newReconciler(t, fakeProber{players: operator.PlayerCount{Online: 3, Max: 20}}, runningServer(), rconSecret())
|
||||
|
||||
s := getServer(t, c, "survival")
|
||||
s.Spec.Idle = v1alpha1.IdleSpec{AutoStopEnabled: true, EmptySecondsBeforeStop: 900}
|
||||
if err := c.Update(context.Background(), s); err != nil {
|
||||
t.Fatalf("enable idle: %v", err)
|
||||
}
|
||||
|
||||
reconcile(t, r, "survival")
|
||||
markPodReady(t, c, "survival")
|
||||
res := reconcile(t, r, "survival")
|
||||
|
||||
if res.RequeueAfter <= 0 {
|
||||
t.Fatalf("RequeueAfter = %v, want a positive probe cadence while occupied", res.RequeueAfter)
|
||||
}
|
||||
}
|
||||
|
||||
// TestNoIdleRequeueWhenDisabled guards against a blanket requeue: servers
|
||||
// without idle auto-stop keep the old quiescent behaviour.
|
||||
func TestNoIdleRequeueWhenDisabled(t *testing.T) {
|
||||
r, c := newReconciler(t, fakeProber{players: operator.PlayerCount{Online: 0, Max: 20}}, runningServer(), rconSecret())
|
||||
|
||||
reconcile(t, r, "survival")
|
||||
markPodReady(t, c, "survival")
|
||||
res := reconcile(t, r, "survival")
|
||||
|
||||
if res.RequeueAfter != 0 {
|
||||
t.Fatalf("RequeueAfter = %v, want 0 when idle auto-stop is disabled", res.RequeueAfter)
|
||||
}
|
||||
}
|
||||
|
||||
// TestIdleAutoStop_SkipsWhenDisabled verifies that a Running empty server does
|
||||
// NOT get an EmptySince timestamp when AutoStopEnabled is false.
|
||||
func TestIdleAutoStop_SkipsWhenDisabled(t *testing.T) {
|
||||
|
||||
Reference in new issue
Block a user