fix(platform): give the control plane a PriorityClass eviction shield (#8)

A full disk made kubelet's node-pressure eviction pick control-plane pods
alongside game pods (both priority 0), and with the images existing only
in the node's containerd (air-gapped), losing the api meant a manual
image re-import. Every control-plane pod template (api/operator/reaper/
registry) now names the bundle's cluster-scoped felis-control-plane
PriorityClass: value 1,000,000, preemptionPolicy Never — eviction order
only, never preempting a running game server. The image-GC half is not
code-fixable on an air-gapped box; troubleshooting gains 13b with the
recovery path (re-run the installer to rebuild imports, or docker save |
k3s ctr images import - for one image).

AUDIT-2026-09-22.md

@@ -17,14 +17,14 @@ IPv6-only 接入（`ssh -6 -i ~/.ssh/id_ed25519 root@fdb2:2c26:f4e4:0:21c:42ff:f
 | 17 | **`ListPendingOpLogins` PG 少列**：接口注释承诺「joined to its staff username」，handler 输出 `username`/`created_at`，fake 正确填充；PG SQL 未 JOIN 也未取 `created_at` → 真机 pending 列表 username 为空、created_at 为 `0001-01-01` | 真机 internal `/op-login/pending` 响应 | `pgrepo.go` SQL 改为 JOIN users + 取 created_at。提交见分支 |
 | 18 | **绑定码并发兑换 500**：`RedeemPlayerBindCode` 无 `FOR UPDATE`（同文件 `VerifyLinkCode` 有），且裸 INSERT。6 路并发同码兑换 → 3×HTTP 500（`users_username_key` 唯一冲突）+ 1×400 + 2×200；跨码并发同 UUID 同样会撞 | 真机四组并发测试 + API 日志 `unmapped error ... duplicate key` | 同码：码行 `FOR UPDATE`（输家干净地 400 invalid_code）；跨码：`INSERT users ... ON CONFLICT (username) DO NOTHING`+重读、`account_links ON CONFLICT (mc_uuid) DO NOTHING`（两路汇聚同一 user）。复测：同码×6=1×200+5×400、双码×2=2×200 同 ID、DB 干净、日志零 unmapped |
 | 19 | **reaper CronJob 渲染位置错误，永远无法调度**：`felis manifests` 把 CronJob 渲染到控制 ns，却引用 minecraft ns 的备份 PVC（Pod 不能跨 ns 挂 PVC：真机 `FailedScheduling: persistentvolumeclaim "felis-backups" not found`）；修正 ns 后又发现 ServerAccount 也不能跨 ns 使用（`serviceaccount "felis-reaper" not found`）。而 reaper 的 Role/RoleBinding 本就在 minecraft ns | 真机三层取证（PVC/SA/调度）| CronJob 与 SA、RoleBinding subject 全部移到 `MinecraftNamespace`（提交 `e4f2cff`+`c839454`）。**修复后完整演练通过**（见下） |
+| 6 | **默认安装无备份能力 + 回收链路不可达/不可用**：(a) 没有任何环节渲染归档 PVC，`FELIS_BACKUP_PVC` 永远为空 → backup/restore 恒 503；(b) bootstrap 从不下传 reaper 三旗标 → 官方安装路径根本无法启用回收；(c) 即便启用，stock k3s 的 `<root>/<pvc>` 布局不存在，且 k3s storage root 是 `0700 root:root`、reaper pod 以 uid 1000 运行 → 真机 `lstat /worlds/…: permission denied`（fail-closed 跳过，但纯空转）；(d) README 口径（“定时备份”）不实 | 真机全链路：默认渲染 PVC+env → 建服→marker→备份 202→jobs 端点 running→succeeded→改 marker→restore→读回原值 ✅；再建 resolvecheck 世界（20d idle，marker）→ CronJob 手动 Job：`evaluated=2 reaped=1`，归档含 marker、PVC+宿主目录回收、`world_backups` 得 `inactive_15d` 行、servers 行/CR 保留 ✅ | `platform/workloads.go` 渲染归档 PVC（minecraft ns、RWO 10Gi、默认 SC），`felis manifests --backup-pvc` 默认 `felis-backups`（`=` 空为显式关闭），bootstrap 统一下传 env/旗标并写 [archive] local_path；`resolveWorldDir` 新增精确 local-path 目录解析（读 PVC `spec.volumeName`，非 glob，杜绝陈旧 PV 目录顶替）；ReaperRole 增 `pvc:get`；bootstrap 设 `FELIS_WORLDS_HOST_PATH` 时给 uid 1000 授 traverse（setfacl/o+x）；README/故障手册改写。提交 `fd33fd0`+`2b87a5a` |
+| 8 | **磁盘打满灾难链**：DiskPressure → kubelet 驱逐控制面（无 PriorityClass 保护）→ 镜像被 GC（无外网、registry 空）→ 全部 ImagePullBackOff；释放后约 8 分钟才恢复调度。恢复依赖人工 `docker save felis:* | k3s ctr images import -`（docker 存储是唯一副本） | 早前真机 drill；本次复核 k3s 默认驱逐顺序 | 控制面（api/operator/reaper/registry）全部挂 `felis-control-plane` PriorityClass（value 1,000,000，`preemptionPolicy: Never`）——节点压力按优先级升序驱逐，游戏 pod（0）先走，控制面留下；镜像 GC 无法用代码根治（air-gap 无回源），新增 troubleshooting §13b 固化恢复路径（重跑安装器重建导入；单镜像可 `docker save | k3s ctr images import -`）。提交见下 |
 
 ## 待决策台账（未修）
 
 | # | 主题 | 说明 |
 |---|------|------|
-| 6 | 默认安装无备份能力 | backup/restore 端点默认 503（需 `FELIS_BACKUP_PVC`+PVC），reaper CronJob 需 `--backup-pvc/--archive-local-path/--worlds-host-path` 三旗标渲染，bootstrap 一个都不传；文档无说明；README 与"自动备份"口径不符。另：归档 3 个月过期依赖 reaper 清理，未启用则磁盘只增不减 |
 | 7 | 异步失败不可感知 | backup/restore 失败后无状态出口：restore 行不变、backup 无行；只有集群侧 Job/日志可查。建议状态字段或 `?failed` 查询 |
-| 8 | 磁盘打满灾难链 | DiskPressure → kubelet 驱逐控制面（无 PriorityClass 保护）→ 镜像被 GC（无外网、registry 空）→ 全部 ImagePullBackOff；释放后约 8 分钟才恢复调度。恢复靠 `docker save felis:* | k3s ctr images import -`（docker 守护进程存储是唯一副本，需固化回源路径）。建议：PriorityClass、镜像入内置 registry、磁盘告警 |
 | 9 | 升级策略 Recreate | 单副本 + Recreate：任何控制面升级=停机；坏升级（实测错 tag）服务中断约 95s 且需人工 `rollout undo`（无自动回滚）。建议 runbook/文档化 |
 | 10 | 备份语义 | 归档包含整个 /data（jar、libraries、cache），167MB；是否符合"world backup"定位待评估 |
 | 11 | PG 断连表现 | 会话查询失败报 401 而非 503（fail-closed 但误导；用户以为没登录） |

docs/troubleshooting.md

@@ -577,6 +577,43 @@ changes, because retention was never conditional in the first place.
 
 ---
 
+## 13b. Node runs out of disk: what survives, and how to recover
+
+A full disk is the one failure this platform cannot ride out by itself, because
+the images exist only in the node's containerd (air-gapped by design), so a
+GC'd image has no pull source.
+
+Eviction ordering. Kubelet's node-pressure eviction removes pods in ascending
+priority. Every control-plane pod (api, operator, reaper, registry) carries the
+bundle's `felis-control-plane` PriorityClass (value 1,000,000,
+`preemptionPolicy: Never`), while game-server pods run at the default 0 — so a
+burst of running servers is evicted first and the control plane keeps serving
+status/console until pressure is genuinely extreme. The class never *preempts*:
+a scheduling decision will not kill a running game server to restart the api.
+
+Symptoms of the image-GC stage: pods stuck `ImagePullBackOff`/`ErrImagePull`
+with `kubectl describe pod` showing a pull attempt for a tag that plainly
+exists (`k3s ctr images ls` will show it missing — the kubelet GC removed it
+under imagefs pressure).
+
+Recovery:
+
+1. Free disk on the node (`df -h /var/lib/rancher`, the biggest consumers are
+   `k3s ctr images ls -q` and the world/backup PVCs under
+   `/var/lib/rancher/k3s/storage`).
+2. Re-import the images by re-running the installer (it rebuilds imports from
+   the local Docker store, which the kubelet GC does not touch):
+   `curl -fsSL <installer URL> | sudo bash` (or `sudo felis setup`), then
+   `kubectl -n felis rollout status deploy/felis-api`.
+3. Delete now-unschedulable stuck pods so they retry with the re-imported image.
+
+If the API itself is down and you only need the images back without a full
+installer run: `docker save felis:<tag> | k3s ctr images import -` restores one
+image from the Docker store (that store is deliberately a second copy; treat it
+as the recovery path, not as free space).
+
+---
+
 ## 14. Metrics for diagnosis (spec §23)
 
 All four mandated metrics have real producers; scrape them when triaging:

internal/platform/bundle.go

@@ -34,7 +34,9 @@ type Object interface {
 // Deployments and the in-cluster registry (Deployment + Service + PVC), which
 // make the SAs and NetworkPolicy peers refer to something real, plus the
 // world-archive PVC that backs backup/restore when a backup PVC is named (see
-// workloads.go).
+// workloads.go). Workloads also carries the one cluster-scoped object, the
+// felis-control-plane PriorityClass (node-pressure eviction shield — a
+// PriorityClass is not namespaced by design); everything else is namespaced.
 // The reaper CronJob is also part of Workloads, rendered only when the retention
 // storage topology is supplied (WorldsHostPath + BackupPVC + ArchiveLocalPath —
 // workloads.go documents the gate and the shape-asserted hostPath caveat). The

internal/platform/workloads.go

@@ -7,6 +7,7 @@ import (
 	appsv1 "k8s.io/api/apps/v1"
 	batchv1 "k8s.io/api/batch/v1"
 	corev1 "k8s.io/api/core/v1"
+	schedulingv1 "k8s.io/api/scheduling/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/util/intstr"
@@ -176,11 +177,14 @@ func InternalAPIBaseURL(controlNamespace string) string {
 // felis-operator Deployment, and the in-cluster registry (Deployment + Service +
 // PVC), the world-archive PVC when p.BackupPVC names it (it backs the
 // backup/restore Jobs and the reaper), plus the reaper CronJob when
-// reaperEnabled(p). FelisImage is required — `felis manifests` enforces it
-// (fail-loud), so a rendered bundle always names a concrete image.
+// reaperEnabled(p). Every pod template carries the felis-control-plane
+// PriorityClass (controlPlanePriorityClass below), the node-pressure eviction
+// shield. FelisImage is required — `felis manifests` enforces it (fail-loud), so
+// a rendered bundle always names a concrete image.
 func Workloads(p Params) []Object {
 	p = p.withDefaults()
 	objs := []Object{
+		controlPlanePriorityClass(),
 		APIDeployment(p),
 		apiService(p),
 		apiInternalService(p),
@@ -199,6 +203,37 @@ func Workloads(p Params) []Object {
 	return objs
 }
 
+// controlPlanePriorityName is the PriorityClass every control-plane workload runs
+// under (api, operator, reaper, registry). Node-pressure eviction (a full disk
+// being the realistic case on a game box) removes pods in ASCENDING priority, and
+// a classless pod is priority 0 — the same as the game servers, which are exactly
+// the pods a burst of joins has just filled the node with. A full disk then takes
+// the api/operator down too, and recovery needs a human: with no reachable
+// registry on an air-gapped box the images are gone, so the fix is a re-run of the
+// installer to re-import them. The class is an eviction shield, not a preemption
+// lever: PreemptionPolicy=Never, so a busy node never loses a running game server
+// merely to schedule the api. Value 1,000,000 sits above every game pod (0) and
+// far below the kubelet's system-reserved classes (2,000,000,000).
+const (
+	controlPlanePriorityName         = "felis-control-plane"
+	controlPlanePriorityValue  int32 = 1_000_000
+	controlPlanePriorityReason       = "Felis control plane (api/operator/reaper/registry) survives node-pressure eviction before game servers"
+)
+
+// controlPlanePriorityClass renders the cluster-scoped PriorityClass the
+// control-plane pod templates reference (the ONE cluster-scoped object in the
+// bundle; a PriorityClass is not namespaced by design).
+func controlPlanePriorityClass() *schedulingv1.PriorityClass {
+	never := corev1.PreemptNever
+	return &schedulingv1.PriorityClass{
+		TypeMeta:         metav1.TypeMeta{APIVersion: "scheduling.k8s.io/v1", Kind: "PriorityClass"},
+		ObjectMeta:       metav1.ObjectMeta{Name: controlPlanePriorityName},
+		Value:            controlPlanePriorityValue,
+		PreemptionPolicy: &never,
+		Description:      controlPlanePriorityReason,
+	}
+}
+
 // reaperEnabled reports whether the retention CronJob should render. It needs all
 // three storage coordinates: WorldsHostPath (where worlds live, mounted to read
 // them), BackupPVC (where archives are written) and ArchiveLocalPath (the mount
@@ -509,6 +544,7 @@ func reaperCronJob(p Params) *batchv1.CronJob {
 						ObjectMeta: metav1.ObjectMeta{Labels: labels},
 						Spec: corev1.PodSpec{
 							ServiceAccountName: SAReaper,
+							PriorityClassName:  controlPlanePriorityName,
 							RestartPolicy:      corev1.RestartPolicyNever,
 							SecurityContext:    hardenedPodSecurityContext(),
 							Containers:         []corev1.Container{container},
@@ -544,6 +580,7 @@ func controlPlaneDeployment(p Params, sa string, container corev1.Container, vol
 				ObjectMeta: metav1.ObjectMeta{Labels: labels},
 				Spec: corev1.PodSpec{
 					ServiceAccountName: sa,
+					PriorityClassName:  controlPlanePriorityName,
 					SecurityContext:    hardenedPodSecurityContext(),
 					Containers:         []corev1.Container{container},
 					Volumes:            volumes,
@@ -591,6 +628,7 @@ func registryDeployment(p Params) *appsv1.Deployment {
 				ObjectMeta: metav1.ObjectMeta{Labels: labels},
 				Spec: corev1.PodSpec{
 					AutomountServiceAccountToken: boolPtr(false),
+					PriorityClassName:            controlPlanePriorityName,
 					SecurityContext:              hardenedPodSecurityContext(),
 					Containers:                   []corev1.Container{container},
 					Volumes: []corev1.Volume{

internal/platform/workloads_test.go

@@ -6,6 +6,7 @@ import (
 	appsv1 "k8s.io/api/apps/v1"
 	batchv1 "k8s.io/api/batch/v1"
 	corev1 "k8s.io/api/core/v1"
+	schedulingv1 "k8s.io/api/scheduling/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/labels"
 )
@@ -459,8 +460,8 @@ func TestRegistry_DeploymentServicePVC(t *testing.T) {
 // internal Service must be present or the login pod's felis-api:8081 path is dead.
 func TestWorkloads_BundleContents(t *testing.T) {
 	objs := Workloads(testParams())
-	if len(objs) != 8 {
-		t.Fatalf("Workloads returned %d objects, want 8", len(objs))
+	if len(objs) != 9 {
+		t.Fatalf("Workloads returned %d objects, want 9", len(objs))
 	}
 	var haveInternalSvc bool
 	for _, o := range objs {
@@ -477,6 +478,59 @@ func TestWorkloads_BundleContents(t *testing.T) {
 	}
 }
 
+// TestControlPlanePriorityClass pins the node-pressure eviction shield: every
+// control-plane pod template (api/operator/reaper/registry) references the one
+// PriorityClass, which outranks the default-0 game pods by eviction order while
+// never preempting them, and is the bundle's only cluster-scoped object. Without
+// this, a full disk evicts the api alongside the game pods and (no reachable
+// registry on an air-gapped box) recovery needs a human re-importing images.
+func TestControlPlanePriorityClass(t *testing.T) {
+	objs := Workloads(reaperParams()) // include the reaper CronJob's pod template
+
+	var class *schedulingv1.PriorityClass
+	clusterScoped := 0
+	deployments, cronJobs := 0, 0
+	for _, o := range objs {
+		if o.GetNamespace() == "" {
+			clusterScoped++
+		}
+		switch v := o.(type) {
+		case *schedulingv1.PriorityClass:
+			class = v
+		case *appsv1.Deployment:
+			deployments++
+			if v.Spec.Template.Spec.PriorityClassName != controlPlanePriorityName {
+				t.Errorf("deployment %s pod priority = %q, want %q", v.Name, v.Spec.Template.Spec.PriorityClassName, controlPlanePriorityName)
+			}
+		case *batchv1.CronJob:
+			cronJobs++
+			if v.Spec.JobTemplate.Spec.Template.Spec.PriorityClassName != controlPlanePriorityName {
+				t.Errorf("CronJob %s pod priority = %q, want %q", v.Name, v.Spec.JobTemplate.Spec.Template.Spec.PriorityClassName, controlPlanePriorityName)
+			}
+		}
+	}
+	if class == nil {
+		t.Fatal("Workloads must render the control-plane PriorityClass")
+	}
+	if class.Name != controlPlanePriorityName || class.Value != controlPlanePriorityValue {
+		t.Errorf("class = %s/%d, want %s/%d", class.Name, class.Value, controlPlanePriorityName, controlPlanePriorityValue)
+	}
+	if class.PreemptionPolicy == nil || *class.PreemptionPolicy != corev1.PreemptNever {
+		t.Errorf("class preemptionPolicy = %v, want Never (an eviction shield, never a lever against running game servers)", class.PreemptionPolicy)
+	}
+	if clusterScoped != 1 {
+		t.Errorf("bundle has %d cluster-scoped objects, want exactly the PriorityClass", clusterScoped)
+	}
+	if deployments != 3 || cronJobs != 1 {
+		t.Errorf("scanned %d deployments / %d cronjobs, want 3 / 1 — a pod template escaped the class check", deployments, cronJobs)
+	}
+	// The class must outrank the default game-pod priority (0); equality would make
+	// the eviction order nondeterministic between the api and a full game server.
+	if controlPlanePriorityValue <= 0 {
+		t.Errorf("control-plane priority %d must exceed the default 0 game-pod priority", controlPlanePriorityValue)
+	}
+}
+
 // cronPodSpec returns the single container and pod template of a CronJob's Job
 // template, failing if the shape is not a single-container pod (the reaper's shape).
 func cronPodSpec(t *testing.T, cj *batchv1.CronJob) (corev1.PodSpec, corev1.Container) {