fix(operator): 服务器停服或删除时丢掉内存里的 RCON 失败计数

internal/operator/autorestart_test.go

@@ -164,6 +164,49 @@ func TestRunningServerDegradesOnlyAfterThreeMisses(t *testing.T) {
 	}
 }
 
+// A miss count lives in the operator's memory, so it has to go with the run it
+// counts: a server meant to stop and a server deleted leave none behind.
+func TestProbeMissCountsGoWithTheServer(t *testing.T) {
+	missOnce := func(t *testing.T) (*operator.Reconciler, client.Client) {
+		t.Helper()
+		down := false
+		r, c := newReconciler(t, switchProber{down: &down}, runningServer(), rconSecret())
+		reconcile(t, r, "survival")
+		markPodReady(t, c, "survival")
+		reconcile(t, r, "survival")
+		down = true
+		reconcile(t, r, "survival")
+		if got := r.ProbeMissesTracked(); got != 1 {
+			t.Fatalf("after a miss: %d servers tracked, want 1", got)
+		}
+		return r, c
+	}
+
+	t.Run("stopped", func(t *testing.T) {
+		r, c := missOnce(t)
+		s := getServer(t, c, "survival")
+		s.Spec.DesiredState = v1alpha1.DesiredStopped
+		if err := c.Update(context.Background(), s); err != nil {
+			t.Fatalf("set desiredState=Stopped: %v", err)
+		}
+		reconcile(t, r, "survival")
+		if got := r.ProbeMissesTracked(); got != 0 {
+			t.Fatalf("after the stop: %d servers tracked, want 0", got)
+		}
+	})
+
+	t.Run("deleted", func(t *testing.T) {
+		r, c := missOnce(t)
+		if err := c.Delete(context.Background(), getServer(t, c, "survival")); err != nil {
+			t.Fatalf("delete server: %v", err)
+		}
+		reconcile(t, r, "survival")
+		if got := r.ProbeMissesTracked(); got != 0 {
+			t.Fatalf("after the delete: %d servers tracked, want 0", got)
+		}
+	})
+}
+
 // A Failed server wakes when its next auto-restart falls due, and every 5m
 // once the attempts are spent, instead of every 2s.
 func TestFailedServerRequeuesWhenTheRetryIsDue(t *testing.T) {

internal/operator/export_test.go

@@ -0,0 +1,9 @@
+package operator
+
+// ProbeMissesTracked is how many servers the reconciler holds a probe-miss count
+// for, so the external tests can see a count go with its server.
+func (r *Reconciler) ProbeMissesTracked() int {
+	r.probeMu.Lock()
+	defer r.probeMu.Unlock()
+	return len(r.probeFailures)
+}

internal/operator/reconciler.go

@@ -143,10 +143,13 @@ type Reconciler struct {
 	Watch *ReconcileWatch
 
 	// probeFailures counts consecutive failed RCON probes of a Running server,
-	// by UID. In memory: an operator restart forgets them, which only delays a
-	// degrade by a few probes.
+	// by namespaced name. In memory: an operator restart forgets them, which only
+	// delays a degrade by a few probes. A count goes when a probe succeeds, when
+	// the server is meant to stop and when it is deleted. A later server of the
+	// same name inherits nothing that matters: it reaches Running only through a
+	// probe that succeeds, and that clears the count.
 	probeMu       sync.Mutex
-	probeFailures map[types.UID]int
+	probeFailures map[types.NamespacedName]int
 }
 
 func (r *Reconciler) now() metav1.Time {
@@ -191,7 +194,11 @@ func (r *Reconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.Resu
 func (r *Reconciler) reconcile(ctx context.Context, req ctrl.Request) (ctrl.Result, error) {
 	var server v1alpha1.MinecraftServer
 	if err := r.Get(ctx, req.NamespacedName, &server); err != nil {
-		// Deletion is handled by owner references on the children.
+		// Deletion is handled by owner references on the children; the probe-miss
+		// count is the one part of a server kept in memory.
+		if apierrors.IsNotFound(err) {
+			r.clearProbeFailures(req.NamespacedName)
+		}
 		return ctrl.Result{}, client.IgnoreNotFound(err)
 	}
 
@@ -208,6 +215,8 @@ func (r *Reconciler) reconcile(ctx context.Context, req ctrl.Request) (ctrl.Resu
 	var res ctrl.Result
 	var err error
 	if desired == v1alpha1.DesiredStopped {
+		// Misses count only against a Running server; the next run starts afresh.
+		r.clearProbeFailures(req.NamespacedName)
 		res, err = r.reconcileStopped(ctx, &server)
 	} else {
 		res, err = r.reconcileRunning(ctx, &server)
@@ -388,7 +397,7 @@ func (r *Reconciler) reconcileRunning(ctx context.Context, server *v1alpha1.Mine
 		if err != nil {
 			// One missed probe of a Running server is noise (a lag spike, a save);
 			// it keeps its status and endpoint until the misses run consecutive.
-			if server.Status.Phase == v1alpha1.PhaseRunning && r.noteProbeFailure(server) < runningProbeMissesBeforeDegrade {
+			if server.Status.Phase == v1alpha1.PhaseRunning && r.noteProbeFailure(client.ObjectKeyFromObject(server)) < runningProbeMissesBeforeDegrade {
 				return ctrl.Result{RequeueAfter: requeueProbeRetry}, nil
 			}
 			r.markStarting(server, "RconNotReachable", err.Error())
@@ -403,7 +412,7 @@ func (r *Reconciler) reconcileRunning(ctx context.Context, server *v1alpha1.Mine
 			}
 			return ctrl.Result{RequeueAfter: r.startingRequeue(server, readinessTimeout(server))}, nil
 		}
-		r.clearProbeFailures(server)
+		r.clearProbeFailures(client.ObjectKeyFromObject(server))
 		players = pc
 		// A tally that could not be read pauses idle auto-stop (below) instead of
 		// counting as an empty server; the condition says so, so a server that
@@ -1060,20 +1069,20 @@ func (r *Reconciler) startingRequeue(server *v1alpha1.MinecraftServer, timeout t
 	return max(wait, requeueStarting)
 }
 
-func (r *Reconciler) noteProbeFailure(server *v1alpha1.MinecraftServer) int {
+func (r *Reconciler) noteProbeFailure(key types.NamespacedName) int {
 	r.probeMu.Lock()
 	defer r.probeMu.Unlock()
 	if r.probeFailures == nil {
-		r.probeFailures = map[types.UID]int{}
+		r.probeFailures = map[types.NamespacedName]int{}
 	}
-	r.probeFailures[server.UID]++
-	return r.probeFailures[server.UID]
+	r.probeFailures[key]++
+	return r.probeFailures[key]
 }
 
-func (r *Reconciler) clearProbeFailures(server *v1alpha1.MinecraftServer) {
+func (r *Reconciler) clearProbeFailures(key types.NamespacedName) {
 	r.probeMu.Lock()
 	defer r.probeMu.Unlock()
-	delete(r.probeFailures, server.UID)
+	delete(r.probeFailures, key)
 }
 
 // recoverFailedStart retries a start that timed out. Once a backoff past the