recovery: skip ignored workers during startup

configs/ananke.tethys.yaml

@@ -282,6 +282,7 @@ startup:
   ignore_workloads: []
   ignore_unavailable_nodes:
     - titan-09
+    - titan-10
   auto_recycle_stuck_pods: true
   auto_quarantine_scheduling_storms: true
   scheduling_storm_event_threshold: 30

configs/ananke.titan-db.yaml

@@ -282,6 +282,7 @@ startup:
   ignore_workloads: []
   ignore_unavailable_nodes:
     - titan-09
+    - titan-10
   auto_recycle_stuck_pods: true
   auto_quarantine_scheduling_storms: true
   scheduling_storm_event_threshold: 30

internal/cluster/orchestrator_scaling.go

@@ -18,13 +18,13 @@ import (
 // Why: keeps behavior explicit so startup/shutdown workflows remain maintainable as services evolve.
 func (o *Orchestrator) effectiveWorkers(ctx context.Context) ([]string, error) {
 	if len(o.cfg.Workers) > 0 {
-		return append([]string{}, o.cfg.Workers...), nil
+		return filterIgnoredNodes(o.cfg.Workers, o.cfg.Startup.IgnoreUnavailableNodes), nil
 	}
 	workers, err := o.discoverWorkers(ctx)
 	if err == nil {
-		return workers, nil
+		return filterIgnoredNodes(workers, o.cfg.Startup.IgnoreUnavailableNodes), nil
 	}
-	fallback := o.fallbackWorkersFromInventory()
+	fallback := filterIgnoredNodes(o.fallbackWorkersFromInventory(), o.cfg.Startup.IgnoreUnavailableNodes)
 	if len(fallback) == 0 {
 		return nil, err
 	}
@@ -32,6 +32,27 @@ func (o *Orchestrator) effectiveWorkers(ctx context.Context) ([]string, error) {
 	return fallback, nil
 }
 
+// filterIgnoredNodes runs one orchestration or CLI step.
+// Signature: filterIgnoredNodes(nodes []string, ignored []string) []string.
+// Why: ignored unavailable nodes should be excluded from active startup and
+// shutdown worker operations, not just readiness gates, so known-absent
+// hardware does not create noisy SSH or uncordon attempts.
+func filterIgnoredNodes(nodes []string, ignored []string) []string {
+	ignoredSet := makeStringSet(ignored)
+	filtered := []string{}
+	for _, node := range nodes {
+		name := strings.TrimSpace(node)
+		if name == "" {
+			continue
+		}
+		if _, skip := ignoredSet[name]; skip {
+			continue
+		}
+		filtered = append(filtered, name)
+	}
+	return filtered
+}
+
 // discoverWorkers runs one orchestration or CLI step.
 // Signature: (o *Orchestrator) discoverWorkers(ctx context.Context) ([]string, error).
 // Why: keeps behavior explicit so startup/shutdown workflows remain maintainable as services evolve.

internal/cluster/orchestrator_unit_additional_test.go

@@ -323,6 +323,28 @@ func TestRecycleStuckControllerPodsHandlesUnknownPodsOnReadyNodes(t *testing.T)
 	}
 }
 
+// TestEffectiveWorkersFiltersIgnoredUnavailableNodes runs one orchestration or CLI step.
+// Signature: TestEffectiveWorkersFiltersIgnoredUnavailableNodes(t *testing.T).
+// Why: ignored unavailable nodes should be excluded before startup tries SSH,
+// k3s-agent start, or uncordon operations against intentionally absent hosts.
+func TestEffectiveWorkersFiltersIgnoredUnavailableNodes(t *testing.T) {
+	cfg := config.Config{
+		Workers: []string{" titan-08 ", "titan-09", "titan-10", "titan-11"},
+		Startup: config.Startup{
+			IgnoreUnavailableNodes: []string{"titan-09", "titan-10"},
+		},
+	}
+	orch := buildOrchestratorWithStubs(t, cfg, nil)
+	got, err := orch.effectiveWorkers(context.Background())
+	if err != nil {
+		t.Fatalf("effectiveWorkers failed: %v", err)
+	}
+	want := []string{"titan-08", "titan-11"}
+	if strings.Join(got, ",") != strings.Join(want, ",") {
+		t.Fatalf("effectiveWorkers mismatch got=%v want=%v", got, want)
+	}
+}
+
 // TestNewConstructsOrchestrator runs one orchestration or CLI step.
 // Signature: TestNewConstructsOrchestrator(t *testing.T).
 // Why: covers constructor path in orchestrator core module.