Merge 部署内置 SPIRE CLI 的 runner 镜像

Reviewed-on: #136

apps/nexus/README.md

@@ -0,0 +1,157 @@
+# Nexus Repository POC
+
+本目录声明一个 Nexus Repository Community Edition POC，用来验证一次性 CI runner 通过
+网络服务复用 Ansible Galaxy、Go Modules 与 OCI/BuildKit 缓存。Nexus 固定为 `3.96.1`，
+镜像固定到官方 multi-arch index digest；LAN 入口为
+`https://nexus.ad.ddupan.top`。
+
+## POC 边界
+
+- 单副本 Deployment，`Recreate` 更新，避免一个 RWO 卷被两个 Pod 同时挂载。
+- `/nexus-data` 使用 `localpv-zfs-ceph` 上的 50 GiB RWO PVC。
+- 资源预算为 250m/2 GiB request、2 CPU/4 GiB limit；JVM heap 上限 2 GiB。
+- 使用容器默认的 embedded H2。它只用于 POC；正式接管 OCI 制品前必须迁移到外部
+  PostgreSQL，并验证备份恢复。
+- 入口只在 LAN wildcard Gateway 上发布，不创建公网 DNS 或 Cloudflare route。
+- 不套 Authelia forward-auth；它会破坏 Go、Ansible 与 OCI 非浏览器客户端。
+- 现有 zot 保持不变。Nexus 完成 OCI、BuildKit cache 和恢复验收前不得迁移或删除 zot。
+
+Terraform provider 创建 `ansible-public`、`go-public`、最小匿名权限与 OCI Bearer Token
+Realm。Nexus 3.94 才加入的原生 OCI repository 已有 REST API，但当前锁定的 community
+provider 尚未暴露 OCI resource；`terraform/reconcile-oci.sh` 因此根据 3.96.1 实例 Swagger
+固定的 JSON schema，幂等调和 `oci-hosted`、`oci-proxy` 与 `oci-public`。不得绕过该入口在
+UI 中创建无人管理的长期 repository。
+
+## 部署
+
+Flux 从 `clusters/homelab/apps/nexus.yaml` 协调本目录，并依赖 Envoy Gateway 与 OpenEBS。
+合并前只渲染配置，不直接 apply：
+
+```bash
+kubectl kustomize apps/nexus
+```
+
+合并并由 Flux 部署后检查：
+
+```bash
+kubectl -n flux-system get kustomization nexus
+kubectl -n nexus get pod,pvc,service,httproute
+kubectl -n nexus logs deployment/nexus --tail=100
+```
+
+启动可能需要数分钟，startup probe 允许最多十分钟。不要因初次启动较慢反复删除 Pod；
+先确认 PVC 已 Bound、Pod 没有 OOM，以及日志仍在推进。
+
+DNS 期望状态已加入 `infrastructure/dns/records.yml`，需从 Samba AD Ansible root 以
+`--check --diff` 核对后再按其 README 应用 DNS tag。没有 DNS 时可先用 port-forward
+验证应用，但不能据此宣称 Gateway 路径已通过。
+
+## 首次初始化与 Terraform
+
+初始管理员密码生成在 PVC 的 `/nexus-data/admin.password`。只在交互式终端中读取并立即
+完成首次密码轮换；不得把密码复制进 shell tracing、工单、Git 或命令参数。随后将
+Terraform 管理账号的凭据存入 OpenBao，由 CI 通过 Terraform input variable 注入以下
+环境变量：
+
+```text
+TF_VAR_nexus_url=https://nexus.ad.ddupan.top
+TF_VAR_nexus_username=admin
+TF_VAR_nexus_password=<OpenBao kv/infra/nexus 的 admin_password 字段>
+```
+
+`terraform/` 使用 `sonatype-nexus-community/sonatyperepo` 1.17.0，当前声明：
+
+- `ansible-galaxy-proxy` → `https://galaxy.ansible.com`
+- `ansible-public` group
+- `go-proxy` → `https://proxy.golang.org`
+- `go-public` group
+
+provider credential 不写入 HCL 或 tfvars。正式 apply 前还必须为这个独立 Terraform root
+配置远端 backend；本地 state 只允许用于可丢弃的 POC，不提交。验证命令：
+
+```bash
+terraform -chdir=apps/nexus/terraform init -backend=false
+terraform -chdir=apps/nexus/terraform validate
+```
+
+先以 `--check` 查看 OCI repository 漂移，再明确 apply；脚本只从上述环境变量取得凭据，
+用临时 `0600` netrc 调用 REST API，退出时删除：
+
+```bash
+apps/nexus/terraform/reconcile-oci.sh --check
+apps/nexus/terraform/reconcile-oci.sh --apply
+```
+
+Terraform 同时把内置 `anonymous` 用户从默认的全仓库 `nx-anonymous` 角色收窄到
+`ansible-public`、其返回制品 URL 使用的 `ansible-galaxy-proxy`、`go-public`，以及
+`oci-public`/`oci-proxy` 的 `browse/read` 权限。`oci-hosted` 不向匿名用户开放。首次接管
+已有实例时先执行
+`terraform import sonatyperepo_user.anonymous anonymous,default`，再 apply；不要先启用默认的
+全仓库匿名读取。
+
+2026-09-20 的 POC 现场验收已确认：Flux 与 Pod Ready、PVC Bound、HTTPRoute 通过 HTTPS
+返回 Nexus 状态 200，Samba DNS 已幂等收敛。全新客户端目录通过匿名入口下载
+`community.general:11.2.0` 时冷缓存为 8.49 秒、热缓存为 1.89 秒，两次 tarball SHA-256
+一致；`golang.org/x/[email protected]` 为 2.92 秒与 1.51 秒。
+
+## 客户端验收
+
+先验证冷缓存，再原样重复命令验证热缓存；记录 Nexus 请求、上游流量和耗时，不只观察
+命令成功。Ansible 配置的 URL 必须以 `/` 结尾：
+
+```ini
+[galaxy]
+server_list = nexus
+
+[galaxy_server.nexus]
+url = https://nexus.ad.ddupan.top/repository/ansible-public/
+```
+
+```bash
+ansible-galaxy collection install -r collections/requirements.yml \
+  -p .ansible/collections
+```
+
+Go POC 使用：
+
+```bash
+GOPROXY=https://nexus.ad.ddupan.top/repository/go-public/ go mod download
+```
+
+私有 `git.ddupan.top/*` module 的 `GOPRIVATE`、凭据与是否允许 `direct` fallback 在实际
+workflow 中单独决定；不要让私有 module path 意外发往公共 proxy。
+
+OCI 使用 path-based routing：匿名公共拉取地址形如
+`nexus.ad.ddupan.top/oci-public/library/alpine:3.22`，认证写入地址形如
+`nexus.ad.ddupan.top/oci-hosted/<namespace>/<image>:<tag>`。2026-09-20 现场验收结果：
+
+- `oci-public` 匿名代理拉取 Alpine 冷缓存 4.75 秒、热缓存 0.80 秒，digest 一致；
+- `oci-hosted` 认证 push/pull 成功，匿名 pull 返回 401；
+- amd64/arm64 OCI image index push 成功，两个平台 manifest 可见；
+- Helm chart push/pull digest 与本地 tarball SHA-256 一致；
+- Cosign 3.1.3 使用一次性密钥签名并验证成功，OCI 1.1 referrers API 返回一个
+  `application/vnd.dev.sigstore.bundle.v0.3+json` artifact；
+- BuildKit `registry` cache 以 `mode=max` 导出成功，销毁首个 builder 后由新 builder 导入，
+  两个 `RUN` step 均明确命中 `CACHED`。
+
+本机安装的测试客户端包括 `/usr/local/bin/cosign` 3.1.3；安装时核对官方 Linux amd64
+binary SHA-256 `4629c757b7618056f8ddd7e2625ae9fdd94c0372a65049520bc7d9df9efc7f71`。
+上述结果仍不代表备份恢复、外部 PostgreSQL 或正式 publisher service account 已完成；
+这些项目通过前不得迁移或删除 zot。
+
+## 数据与恢复
+
+POC 的数据库、配置、blob、初始管理员状态都位于 `nexus-data` PVC。删除 Deployment
+不会删除 PVC；删除 PVC 会永久删除整个 POC。当前没有独立备份，不能将它用于唯一副本的
+正式制品。
+
+恢复验证至少包括：停止写入、取得一致备份、在独立 PVC/实例恢复、登录、列出 repository、
+拉取已缓存的 Ansible/Go 制品，并核对 OCI digest/referrers。正式化时再把数据库迁移至
+外部 PostgreSQL，并分别定义数据库与 blob 的备份、恢复顺序和 RPO。
+
+参考：
+
+- [Nexus OCI repositories](https://help.sonatype.com/en/oci-repositories.html)
+- [Nexus Ansible repositories](https://help.sonatype.com/en/ansible-repositories.html)
+- [Nexus Go repositories](https://help.sonatype.com/en/go-repositories.html)
+- [官方容器镜像](https://hub.docker.com/r/sonatype/nexus3)

apps/nexus/deployment.yaml

@@ -0,0 +1,82 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: nexus
+  namespace: nexus
+  labels:
+    app.kubernetes.io/name: nexus
+spec:
+  replicas: 1
+  strategy:
+    type: Recreate
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: nexus
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: nexus
+    spec:
+      automountServiceAccountToken: false
+      securityContext:
+        fsGroup: 200
+        fsGroupChangePolicy: OnRootMismatch
+        runAsGroup: 200
+        runAsNonRoot: true
+        runAsUser: 200
+        seccompProfile:
+          type: RuntimeDefault
+      terminationGracePeriodSeconds: 120
+      containers:
+        - name: nexus
+          image: docker.io/sonatype/nexus3:3.96.1@sha256:56142f13432cf072e017aebb2025f201e42ae36ff40bb82618c702504c61f7dd
+          imagePullPolicy: IfNotPresent
+          env:
+            - name: INSTALL4J_ADD_VM_PARAMS
+              value: >-
+                -Xms1024m -Xmx2048m -XX:MaxDirectMemorySize=1024m
+                -Djava.util.prefs.userRoot=/nexus-data/javaprefs
+          ports:
+            - name: http
+              containerPort: 8081
+              protocol: TCP
+          resources:
+            requests:
+              cpu: 250m
+              memory: 2Gi
+            limits:
+              cpu: "2"
+              memory: 4Gi
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop:
+                - ALL
+          startupProbe:
+            httpGet:
+              path: /service/rest/v1/status
+              port: http
+            failureThreshold: 60
+            periodSeconds: 10
+            timeoutSeconds: 5
+          readinessProbe:
+            httpGet:
+              path: /service/rest/v1/status
+              port: http
+            failureThreshold: 6
+            periodSeconds: 10
+            timeoutSeconds: 5
+          livenessProbe:
+            httpGet:
+              path: /service/rest/v1/status
+              port: http
+            failureThreshold: 6
+            periodSeconds: 30
+            timeoutSeconds: 5
+          volumeMounts:
+            - name: data
+              mountPath: /nexus-data
+      volumes:
+        - name: data
+          persistentVolumeClaim:
+            claimName: nexus-data

apps/nexus/httproute.yaml

@@ -0,0 +1,16 @@
+apiVersion: gateway.networking.k8s.io/v1
+kind: HTTPRoute
+metadata:
+  name: nexus
+  namespace: nexus
+spec:
+  parentRefs:
+    - name: eg
+      namespace: envoy-gateway-system
+      sectionName: https
+  hostnames:
+    - nexus.ad.ddupan.top
+  rules:
+    - backendRefs:
+        - name: nexus
+          port: 8081

apps/nexus/kustomization.yaml

@@ -0,0 +1,9 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - namespace.yaml
+  - pvc.yaml
+  - deployment.yaml
+  - service.yaml
+  - httproute.yaml
+  - networkpolicy.yaml

apps/nexus/namespace.yaml

@@ -0,0 +1,8 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: nexus
+  labels:
+    pod-security.kubernetes.io/enforce: restricted
+    pod-security.kubernetes.io/audit: restricted
+    pod-security.kubernetes.io/warn: restricted

apps/nexus/networkpolicy.yaml

@@ -0,0 +1,23 @@
+apiVersion: networking.k8s.io/v1
+kind: NetworkPolicy
+metadata:
+  name: nexus-ingress
+  namespace: nexus
+spec:
+  podSelector:
+    matchLabels:
+      app.kubernetes.io/name: nexus
+  policyTypes:
+    - Ingress
+  ingress:
+    - from:
+        - namespaceSelector:
+            matchLabels:
+              kubernetes.io/metadata.name: envoy-gateway-system
+          podSelector:
+            matchLabels:
+              gateway.envoyproxy.io/owning-gateway-name: eg
+              gateway.envoyproxy.io/owning-gateway-namespace: envoy-gateway-system
+      ports:
+        - protocol: TCP
+          port: 8081

apps/nexus/pvc.yaml

@@ -0,0 +1,12 @@
+apiVersion: v1
+kind: PersistentVolumeClaim
+metadata:
+  name: nexus-data
+  namespace: nexus
+spec:
+  accessModes:
+    - ReadWriteOnce
+  resources:
+    requests:
+      storage: 50Gi
+  storageClassName: localpv-zfs-ceph

apps/nexus/service.yaml

@@ -0,0 +1,14 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: nexus
+  namespace: nexus
+spec:
+  type: ClusterIP
+  selector:
+    app.kubernetes.io/name: nexus
+  ports:
+    - name: http
+      port: 8081
+      protocol: TCP
+      targetPort: http

apps/nexus/terraform/.gitignore

@@ -0,0 +1,6 @@
+.terraform/
+*.tfstate
+*.tfstate.*
+*.tfplan
+crash.log
+crash.*.log

apps/nexus/terraform/.terraform.lock.hcl

@@ -0,0 +1,24 @@
+# This file is maintained automatically by "terraform init".
+# Manual edits may be lost in future updates.
+
+provider "registry.terraform.io/sonatype-nexus-community/sonatyperepo" {
+  version     = "1.17.0"
+  constraints = "1.17.0"
+  hashes = [
+    "h1:uKhvhhhI7B+HBsh0zq/ybqKt+EnOGyI6rjcRCtj79ZA=",
+    "zh:0dde99e7b343fa01f8eefc378171fb8621bedb20f59157d6cc8e3d46c738105f",
+    "zh:3315929df254a3a6ac27c8c846c2006f7d2a91fadc014351bc4d617f948e5bf9",
+    "zh:36be5a455af3ce4e187de26753de63e78c1ee9a32dba0135c6cf96a6c1fff25f",
+    "zh:3f73f7ff57b8c339a7c7ac37653e2dc0b2dd9dcc3f3a538788e7e3ac838337b2",
+    "zh:40286ecca4c22ab7ae90618ac6d2743f5055199dac81cf5204a4a397c784d439",
+    "zh:4d24e5c0195fb3155b1967583ee64cfeda402d7cc7f3c73369438f6c69f4245b",
+    "zh:828a9d7aceaac36af7f9c07af43ec8d20a89148780645d170ffb1c68b2da792d",
+    "zh:a5ab04de3fe626ec57c832618c6f990abd6610f81e132621651e0b180b970cff",
+    "zh:a959fa6090a8c0f53739879184e7346423494aee598003df0d1ab4a22b2eee91",
+    "zh:bdda26c2f03f918bbe59e75abea44868fafda019c3a543725331195df126350b",
+    "zh:d8048e149ee97ba62971e6a79355d59887bc6d10fcf72cc2feff3d0a2582670c",
+    "zh:dd36f9988af4e1ca5b1ca7b7bb6f658df9a220dfcda7fec7392fedfe9064f652",
+    "zh:dda2688d46c7e539fe97e8fe9d3ec81fb364170e018d9c6a681364c8955d4e9d",
+    "zh:e6b519afe2dea1c0434f766eb6bc9ba78cc5b6ef2c311c2ca3c65cb24744f31f",
+  ]
+}

apps/nexus/terraform/oci/oci-hosted.json

@@ -0,0 +1,17 @@
+{
+  "name": "oci-hosted",
+  "online": true,
+  "storage": {
+    "blobStoreName": "default",
+    "strictContentTypeValidation": true,
+    "writePolicy": "ALLOW",
+    "latestPolicy": false
+  },
+  "oci": {
+    "v1Enabled": false,
+    "forceBasicAuth": false,
+    "pathEnabled": true
+  },
+  "component": { "proprietaryComponents": false },
+  "cosign": { "enforcement": "NONE" }
+}

apps/nexus/terraform/oci/oci-proxy.json

@@ -0,0 +1,27 @@
+{
+  "name": "oci-proxy",
+  "online": true,
+  "storage": {
+    "blobStoreName": "default",
+    "strictContentTypeValidation": true
+  },
+  "oci": {
+    "v1Enabled": false,
+    "forceBasicAuth": false,
+    "pathEnabled": true
+  },
+  "ociProxy": {
+    "indexType": "HUB",
+    "cacheForeignLayers": false,
+    "foreignLayerUrlWhitelist": []
+  },
+  "proxy": {
+    "remoteUrl": "https://registry-1.docker.io",
+    "contentMaxAge": 1440,
+    "metadataMaxAge": 60,
+    "preserveEncodedCharacters": false
+  },
+  "negativeCache": { "enabled": true, "timeToLive": 60 },
+  "httpClient": { "blocked": false, "autoBlock": true },
+  "cosign": { "enforcement": "NONE" }
+}

apps/nexus/terraform/oci/oci-public.json

@@ -0,0 +1,15 @@
+{
+  "name": "oci-public",
+  "online": true,
+  "storage": {
+    "blobStoreName": "default",
+    "strictContentTypeValidation": true
+  },
+  "group": { "memberNames": ["oci-proxy"] },
+  "oci": {
+    "v1Enabled": false,
+    "forceBasicAuth": false,
+    "pathEnabled": true
+  },
+  "cosign": { "enforcement": "NONE" }
+}

apps/nexus/terraform/reconcile-oci.sh

@@ -0,0 +1,80 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+mode="${1:---check}"
+case "$mode" in
+  --check | --apply) ;;
+  *) echo "usage: $0 [--check|--apply]" >&2; exit 2 ;;
+esac
+
+: "${TF_VAR_nexus_url:?set TF_VAR_nexus_url}"
+: "${TF_VAR_nexus_username:?set TF_VAR_nexus_username}"
+: "${TF_VAR_nexus_password:?set TF_VAR_nexus_password}"
+
+script_dir="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)"
+auth_file="$(mktemp /tmp/nexus-oci-auth.XXXXXX)"
+trap 'rm -f -- "$auth_file"' EXIT
+chmod 0600 "$auth_file"
+printf 'machine %s\nlogin %s\npassword %s\n' \
+  "${TF_VAR_nexus_url#*://}" "$TF_VAR_nexus_username" \
+  "$TF_VAR_nexus_password" >"$auth_file"
+
+drift=0
+for entry in \
+  "hosted:$script_dir/oci/oci-hosted.json" \
+  "proxy:$script_dir/oci/oci-proxy.json" \
+  "group:$script_dir/oci/oci-public.json"; do
+  repository_type="${entry%%:*}"
+  desired_file="${entry#*:}"
+  repository_name="$(jq -er '.name' "$desired_file")"
+  endpoint="$TF_VAR_nexus_url/service/rest/v1/repositories/oci/$repository_type"
+  current_file="$(mktemp /tmp/nexus-oci-current.XXXXXX)"
+
+  status="$(curl --silent --show-error --netrc-file "$auth_file" \
+    --output "$current_file" --write-out '%{http_code}' \
+    "$endpoint/$repository_name")"
+
+  if [[ "$status" == 404 ]]; then
+    drift=1
+    if [[ "$mode" == --apply ]]; then
+      curl --fail --silent --show-error --netrc-file "$auth_file" \
+        --header 'Content-Type: application/json' \
+        --data-binary "@$desired_file" "$endpoint"
+      echo "created $repository_name"
+    else
+      echo "missing $repository_name" >&2
+    fi
+  elif [[ "$status" == 200 ]]; then
+    if jq -e --slurpfile desired "$desired_file" '
+      def subset($actual; $wanted):
+        if ($wanted | type) == "object" then
+          all($wanted | keys[];
+            ($actual[.] != null) and subset($actual[.]; $wanted[.]))
+        else
+          $actual == $wanted
+        end;
+      subset(.; $desired[0])
+    ' "$current_file" >/dev/null; then
+      echo "in sync $repository_name"
+    else
+      drift=1
+      if [[ "$mode" == --apply ]]; then
+        curl --fail --silent --show-error --netrc-file "$auth_file" \
+          --request PUT --header 'Content-Type: application/json' \
+          --data-binary "@$desired_file" "$endpoint/$repository_name"
+        echo "updated $repository_name"
+      else
+        echo "drifted $repository_name" >&2
+      fi
+    fi
+  else
+    cat "$current_file" >&2
+    echo "unexpected HTTP $status for $repository_name" >&2
+    exit 1
+  fi
+  rm -f -- "$current_file"
+done
+
+if [[ "$mode" == --check && "$drift" -ne 0 ]]; then
+  exit 1
+fi

apps/nexus/terraform/repositories.tf

@@ -0,0 +1,64 @@
+locals {
+  proxy_http_client = {
+    auto_block = true
+    blocked    = false
+  }
+
+  proxy_negative_cache = {
+    enabled      = true
+    time_to_live = 60
+  }
+
+  repository_storage = {
+    blob_store_name                = "default"
+    strict_content_type_validation = true
+  }
+}
+
+resource "sonatyperepo_repository_ansiblegalaxy_proxy" "galaxy" {
+  name   = "ansible-galaxy-proxy"
+  online = true
+
+  http_client    = local.proxy_http_client
+  negative_cache = local.proxy_negative_cache
+  proxy = {
+    remote_url       = "https://galaxy.ansible.com"
+    content_max_age  = 1440
+    metadata_max_age = 60
+  }
+  storage = local.repository_storage
+}
+
+resource "sonatyperepo_repository_ansiblegalaxy_group" "public" {
+  name   = "ansible-public"
+  online = true
+
+  group = {
+    member_names = [sonatyperepo_repository_ansiblegalaxy_proxy.galaxy.name]
+  }
+  storage = local.repository_storage
+}
+
+resource "sonatyperepo_repository_go_proxy" "public" {
+  name   = "go-proxy"
+  online = true
+
+  http_client    = local.proxy_http_client
+  negative_cache = local.proxy_negative_cache
+  proxy = {
+    remote_url       = "https://proxy.golang.org"
+    content_max_age  = 1440
+    metadata_max_age = 60
+  }
+  storage = local.repository_storage
+}
+
+resource "sonatyperepo_repository_go_group" "public" {
+  name   = "go-public"
+  online = true
+
+  group = {
+    member_names = [sonatyperepo_repository_go_proxy.public.name]
+  }
+  storage = local.repository_storage
+}

apps/nexus/terraform/security.tf

@@ -0,0 +1,75 @@
+resource "sonatyperepo_privilege_repository_view" "anonymous_ansible" {
+  name        = "ci-anonymous-ansible-read"
+  description = "Anonymous read access to the Ansible Galaxy group"
+  actions     = ["BROWSE", "READ"]
+  format      = "ansiblegalaxy"
+  repository  = sonatyperepo_repository_ansiblegalaxy_group.public.name
+}
+
+resource "sonatyperepo_privilege_repository_view" "anonymous_ansible_proxy" {
+  name        = "ci-anonymous-ansible-proxy-read"
+  description = "Anonymous artifact read access to the Ansible Galaxy proxy"
+  actions     = ["BROWSE", "READ"]
+  format      = "ansiblegalaxy"
+  repository  = sonatyperepo_repository_ansiblegalaxy_proxy.galaxy.name
+}
+
+resource "sonatyperepo_privilege_repository_view" "anonymous_go" {
+  name        = "ci-anonymous-go-read"
+  description = "Anonymous read access to the Go module group"
+  actions     = ["BROWSE", "READ"]
+  format      = "go"
+  repository  = sonatyperepo_repository_go_group.public.name
+}
+
+resource "sonatyperepo_privilege_repository_view" "anonymous_oci_public" {
+  name        = "ci-anonymous-oci-public-read"
+  description = "Anonymous read access to the public OCI group"
+  actions     = ["BROWSE", "READ"]
+  format      = "oci"
+  repository  = "oci-public"
+}
+
+resource "sonatyperepo_privilege_repository_view" "anonymous_oci_proxy" {
+  name        = "ci-anonymous-oci-proxy-read"
+  description = "Anonymous read access to the OCI proxy member"
+  actions     = ["BROWSE", "READ"]
+  format      = "oci"
+  repository  = "oci-proxy"
+}
+
+resource "sonatyperepo_role" "anonymous_ci" {
+  id          = "ci-anonymous-read"
+  name        = "CI anonymous read"
+  description = "Read-only access to public CI dependency proxy groups"
+  privileges = [
+    sonatyperepo_privilege_repository_view.anonymous_ansible.name,
+    sonatyperepo_privilege_repository_view.anonymous_ansible_proxy.name,
+    sonatyperepo_privilege_repository_view.anonymous_go.name,
+    sonatyperepo_privilege_repository_view.anonymous_oci_public.name,
+    sonatyperepo_privilege_repository_view.anonymous_oci_proxy.name,
+  ]
+  roles = []
+}
+
+resource "sonatyperepo_user" "anonymous" {
+  user_id       = "anonymous"
+  first_name    = "Anonymous"
+  last_name     = "User"
+  email_address = "[email protected]"
+  status        = "active"
+  roles         = [sonatyperepo_role.anonymous_ci.id]
+}
+
+resource "sonatyperepo_system_anonymous_access" "ci" {
+  enabled    = true
+  user_id    = sonatyperepo_user.anonymous.user_id
+  realm_name = "NexusAuthorizingRealm"
+}
+
+resource "sonatyperepo_security_realms" "active" {
+  active = [
+    "NexusAuthenticatingRealm",
+    "OciBearerToken",
+  ]
+}

apps/nexus/terraform/versions.tf

@@ -0,0 +1,33 @@
+terraform {
+  required_version = ">= 1.11.0"
+
+  required_providers {
+    sonatyperepo = {
+      source  = "sonatype-nexus-community/sonatyperepo"
+      version = "1.17.0"
+    }
+  }
+}
+
+provider "sonatyperepo" {
+  url      = var.nexus_url
+  username = var.nexus_username
+  password = var.nexus_password
+}
+
+variable "nexus_url" {
+  description = "Nexus Repository base URL"
+  type        = string
+}
+
+variable "nexus_username" {
+  description = "Nexus Terraform management username"
+  type        = string
+  sensitive   = true
+}
+
+variable "nexus_password" {
+  description = "Nexus Terraform management password"
+  type        = string
+  sensitive   = true
+}

clusters/homelab/README.md

@@ -53,4 +53,6 @@ sudo k3s kubectl -n flux-system get gitrepositories,kustomizations
 - SPIRE 已按 hardened chart 内部 fork `0.30.2-ddupan.1`（基于上游 `0.30.2`，SPIRE
   `1.15.3`）声明，使用共享
   PostgreSQL 与独立 signing-key PVC；首次上线和 OpenBao JWT-SVID PoC 尚待合并后验证；
+- Nexus Repository CE POC 已加入 GitOps 声明，计划验证 Ansible Galaxy、Go Modules 与
+  OCI/BuildKit 缓存；尚未部署或完成现场验收，现有 zot 保持不变；
 - root Kustomization 与所有 brownfield 子 Kustomization 继续保持 `prune: false`。

clusters/homelab/apps/dynamic-runner.yaml

@@ -6,11 +6,12 @@ metadata:
 spec:
   dependsOn:
     - name: external-secrets
-    - name: nats
     - name: spire
   interval: 10m
   path: ./platform/dynamic-runner
-  prune: false
+  # The runner backends are replaceable. Prune is required when a retired
+  # worker is removed from the component; otherwise it keeps consuming work.
+  prune: true
   sourceRef:
     kind: GitRepository
     name: flux-system

clusters/homelab/apps/nexus.yaml

@@ -0,0 +1,22 @@
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: nexus
+  namespace: flux-system
+spec:
+  dependsOn:
+    - name: envoy-gateway
+    - name: openebs
+  healthChecks:
+    - apiVersion: apps/v1
+      kind: Deployment
+      name: nexus
+      namespace: nexus
+  interval: 10m
+  path: ./apps/nexus
+  prune: false
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  timeout: 15m
+  wait: true

clusters/homelab/kustomization.yaml

@@ -7,7 +7,6 @@ resources:
   - apps/envoy-gateway.yaml
   - apps/external-secrets.yaml
   - apps/gitea.yaml
-  - apps/gitea-actions.yaml
   - apps/http-echo.yaml
   - apps/openebs.yaml
   - apps/nats.yaml
@@ -15,3 +14,4 @@ resources:
   - apps/spire.yaml
   - apps/observability.yaml
   - apps/zot.yaml
+  - apps/nexus.yaml

clusters/sandbox/README.md

@@ -15,7 +15,8 @@ Root bootstrap 已完成。后续按依赖顺序分别引入：
 1. 监控 CRD、kube-state-metrics 以及 kubelet/cAdvisor 抓取配置；
 2. SPIRE Agent、SPIFFE CSI Driver 与 workload registration；
 3. Kata Containers、`block-plain` RuntimeClass；
-4. OpenSandbox operator/server 及 `ci-pod`、`ci-vm` Pools。
+4. 独立 External Secrets Operator 与 sandbox 专用 OpenBao auth backend；
+5. OpenSandbox controller/server；CI Pool 与 runner 调度器随后独立接入。
 
 每一阶段单独合并并等待对应 Flux Kustomization Ready，不在 bootstrap 时一次性部署。
 第一阶段监控拆为 `monitoring-operator` 与依赖它的 `monitoring`，防止 VM CR 在
@@ -42,15 +43,25 @@ attestation。Server 使用 external bundle publisher 持续维护 sandbox
 显式关闭 Server 与 OIDC Provider，只部署 Agent DaemonSet 和 SPIFFE CSI Driver；因此
 不会产生第二个 trust root。
 
-`spire-smoke` namespace、ServiceAccount 和 `sandbox-spire-smoke` ClusterSPIFFEID 是
+`spire-smoke` namespace、ServiceAccount 和 `sandbox-spire-smoke` ClusterSPIFFEID 只用于
-普通 Pod 与后续 Kata guest 的回归夹具，稳定身份为
+普通 Pod 的 CSI 回归夹具，稳定身份为 `spiffe://ddupan.top/sandbox/smoke`。Kata guest
-`spiffe://ddupan.top/sandbox/smoke`。测试 Pod 临时创建并在验收后删除，身份声明保留。
+不能复用 node Agent 暴露的 Unix socket；virtio-fs 只能呈现 socket 路径，不能把连接
+跨过 VM 边界。Kata workload 必须使用 guest 内 Agent，具体约束见
+`platform/sandbox-kata/README.md`。测试 Pod 临时创建并在验收后删除，普通 Pod 的身份
+声明保留。
 
 Kata 阶段使用官方 4.1.0 `kata-deploy` chart 的短生命周期 `job` 模式，逐节点安装并
 重启 K3s。只启用 `kata-clh-runtime-rs`，不创建默认 `kata` 别名；该 handler 的
 `emptyDir` 固定使用 `block-plain`，为 Docker/BuildKit overlay2 与 kind 提供 guest
 内块设备文件系统。详细限制与上线验收见 `platform/sandbox-kata/README.md`。
 
+Sandbox 的 ESO 通过独立 `auth/kubernetes-sandbox` 向 OpenBao 证明 ServiceAccount
+身份，只能读取共享的 `kv/k8s/opensandbox-api`。它不保存 reviewer JWT 或长期 Bao token；相关
+Terraform 与 Flux 边界见 `platform/sandbox-external-secrets/README.md`。
+
+OpenSandbox 阶段固定官方源码 commit 与 umbrella chart `0.2.2`，只部署 controller、
+ClusterIP server 和 CRD。API key 由 ESO 从 OpenBao 投影，明文不进入 Git。
+
 ## 监控边界
 
 这里只管理 sandbox LXC 内的 Kubernetes 监控，不负责 PVE 宿主监控。LXC 与宿主共享

clusters/sandbox/apps/ci-runners.yaml

@@ -2,13 +2,14 @@
 apiVersion: kustomize.toolkit.fluxcd.io/v1
 kind: Kustomization
 metadata:
-  name: opensandbox-pools
+  name: ci-runners
   namespace: flux-system
 spec:
   dependsOn:
     - name: opensandbox
+    - name: spire-agents
   interval: 10m
-  path: ./platform/sandbox-opensandbox-pools
+  path: ./platform/sandbox-ci-runners
   prune: true
   sourceRef:
     kind: GitRepository

clusters/sandbox/apps/external-secrets-operator.yaml

@@ -0,0 +1,17 @@
+---
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: external-secrets-operator
+  namespace: flux-system
+spec:
+  dependsOn:
+    - name: monitoring-operator
+  interval: 10m
+  path: ./platform/sandbox-external-secrets/operator
+  prune: true
+  sourceRef:
+    kind: GitRepository
+    name: flux-system
+  timeout: 10m
+  wait: true

clusters/sandbox/apps/external-secrets.yaml

@@ -1,14 +1,17 @@
+---
 apiVersion: kustomize.toolkit.fluxcd.io/v1
 kind: Kustomization
 metadata:
-  name: gitea-actions
+  name: external-secrets
   namespace: flux-system
 spec:
+  dependsOn:
+    - name: external-secrets-operator
   interval: 10m
-  path: ./platform/gitea-runner
+  path: ./platform/sandbox-external-secrets/config
-  prune: false
+  prune: true
   sourceRef:
     kind: GitRepository
     name: flux-system
-  timeout: 3m
+  timeout: 10m
-  wait: false
+  wait: true

clusters/sandbox/apps/opensandbox.yaml

@@ -6,13 +6,14 @@ metadata:
   namespace: flux-system
 spec:
   dependsOn:
+    - name: external-secrets
     - name: kata
-    - name: spire-agents
+    - name: monitoring-operator
   interval: 10m
   path: ./platform/sandbox-opensandbox
   prune: true
   sourceRef:
     kind: GitRepository
     name: flux-system
-  timeout: 20m
+  timeout: 15m
   wait: true

clusters/sandbox/kustomization.yaml

@@ -7,5 +7,7 @@ resources:
   - apps/spire-bootstrap.yaml
   - apps/spire-agents.yaml
   - apps/kata.yaml
+  - apps/external-secrets-operator.yaml
+  - apps/external-secrets.yaml
   - apps/opensandbox.yaml
-  - apps/opensandbox-pools.yaml
+  - apps/ci-runners.yaml

infrastructure/dns/records.yml

@@ -18,6 +18,7 @@ homelab_dns:
       - { zone: ad.ddupan.top, name: metrics-write, type: A, values: [192.168.10.127] }
       - { zone: ad.ddupan.top, name: netbox, type: A, values: [192.168.10.127] }
       - { zone: ad.ddupan.top, name: nats, type: A, values: [192.168.10.127] }
+      - { zone: ad.ddupan.top, name: nexus, type: A, values: [192.168.10.127] }
       - { zone: ad.ddupan.top, name: s3, type: A, values: [192.168.10.127] }
       - { zone: ad.ddupan.top, name: spire-oidc, type: A, values: [192.168.10.127] }
       - { zone: ad.ddupan.top, name: spire-server, type: A, values: [192.168.10.127] }

infrastructure/openbao/README.md

@@ -51,7 +51,7 @@ few things Terraform must not own.**
 | secrets engine mounts (`kv`, `pki`, `ssh-client-signer`) | the daemon, Raft, TLS files, systemd |
 | PKI role, issuing/CRL URLs, cluster paths, **ACME** | `bao operator init` / unseal (manual, PGP-wrapped) |
 | SSH signing role (`ai-agent`) | **PKI root CA + SSH CA signing key** |
-| OIDC auth *mount* and *role* | OIDC **client secret** (`auth/oidc/config`) |
+| OIDC/Kubernetes auth mount、config 与 role | OIDC **client secret** (`auth/oidc/config`) |
 | all policies | snapshot token + script + systemd timer |
 |  | host-level CA trust distribution (`openbao_ssh_ca_trust`) |
 
@@ -92,6 +92,13 @@ then `VAULT_ADDR`/`VAULT_TOKEN`), mirroring how `smtp-relay/terraform` uses
 native `openbao/openbao` provider is published only to the OpenTofu registry and
 cannot be resolved by the HashiCorp `terraform` CLI.
 
+Sandbox 集群使用独立的 `auth/kubernetes-sandbox`。其 API 地址、公开 Kubernetes CA、
+ESO role 与只读 `kv/k8s/opensandbox-api` policy 全部由 Terraform 管理；CA 位于
+`terraform/certs/sandbox-kubernetes-ca.crt`。集群重建并轮换 CA 后，先更新该文件并
+apply，再让 Flux 恢复 ESO reconciliation。该 backend 不保存 reviewer JWT，而是使用
+ESO 的短期登录 JWT 执行 TokenReview。该组资源已于 2026-09-18 apply，随后复验 plan
+为 zero-diff。
+
 ## DNS
 
 `bao.ad.ddupan.top` is an **internal-only** name — not a public Cloudflare record and

infrastructure/openbao/terraform/auth.tf

@@ -59,3 +59,33 @@ resource "vault_kubernetes_auth_backend_role" "external_secrets" {
   # in a long TTL and every extra hour is a longer-lived credential in memory.
   token_ttl = 3600
 }
+
+# A Kubernetes auth mount can validate identities from only the API server it is
+# configured against. The sandbox cluster therefore cannot reuse auth/kubernetes,
+# whose TokenReview endpoint belongs to homelab.
+resource "vault_auth_backend" "sandbox_kubernetes" {
+  type = "kubernetes"
+  path = "kubernetes-sandbox"
+}
+
+resource "vault_kubernetes_auth_backend_config" "sandbox" {
+  backend              = vault_auth_backend.sandbox_kubernetes.path
+  kubernetes_host      = "https://10.60.0.13:6443"
+  kubernetes_ca_cert   = file("${path.module}/certs/sandbox-kubernetes-ca.crt")
+  disable_local_ca_jwt = true
+
+  # Deliberately omit token_reviewer_jwt. OpenBao uses the login JWT for
+  # TokenReview; the sandbox external-secrets ServiceAccount is bound only to
+  # system:auth-delegator and all issued JWTs remain short-lived.
+}
+
+resource "vault_kubernetes_auth_backend_role" "sandbox_external_secrets" {
+  backend   = vault_auth_backend.sandbox_kubernetes.path
+  role_name = "external-secrets"
+
+  bound_service_account_names      = ["external-secrets"]
+  bound_service_account_namespaces = ["external-secrets"]
+
+  token_policies = [vault_policy.sandbox_external_secrets.name]
+  token_ttl      = 3600
+}

infrastructure/openbao/terraform/certs/sandbox-kubernetes-ca.crt

@@ -0,0 +1,10 @@
+-----BEGIN CERTIFICATE-----
+MIIBdjCCAR2gAwIBAgIBADAKBggqhkjOPQQDAjAjMSEwHwYDVQQDDBhrM3Mtc2Vy
+dmVyLWNhQDE3ODk2NTI4NTcwHhcNMjYwOTE3MTI0NzM3WhcNMzYwOTE0MTI0NzM3
+WjAjMSEwHwYDVQQDDBhrM3Mtc2VydmVyLWNhQDE3ODk2NTI4NTcwWTATBgcqhkjO
+PQIBBggqhkjOPQMBBwNCAAR4SbqzTXZnlZdUPz7viN6+dYbB1Maw44Qiepn9r5XG
+sOzkYkN8t1aG3Ugo8TqQ3xJaKkM89n1Rluj0vbOhiNajo0IwQDAOBgNVHQ8BAf8E
+BAMCAqQwDwYDVR0TAQH/BAUwAwEB/zAdBgNVHQ4EFgQUyLSGoKAKAJuiniuRdBLG
+XYaDQC8wCgYIKoZIzj0EAwIDRwAwRAIgFkVzyUZexk/ynnxBEOg+3foJv3WKqAei
+hTSRjO1gL0UCIFbBKR7BMrJJAgW3DJFeeBM+b+tTg93jNx55qZACbFOL
+-----END CERTIFICATE-----

infrastructure/openbao/terraform/policies.tf

@@ -35,3 +35,10 @@ resource "vault_policy" "external_secrets" {
   name   = "external-secrets"
   policy = file("${path.module}/policies/external-secrets.hcl")
 }
+
+# The sandbox cluster has its own Kubernetes auth backend and a deliberately
+# narrower KV view than the homelab ESO instance.
+resource "vault_policy" "sandbox_external_secrets" {
+  name   = "sandbox-external-secrets"
+  policy = file("${path.module}/policies/sandbox-external-secrets.hcl")
+}

infrastructure/openbao/terraform/policies/sandbox-external-secrets.hcl

@@ -0,0 +1,9 @@
+# Read only the shared OpenSandbox control-plane API key. The same Bao object is
+# consumed by the server in sandbox and, later, by the scheduler in homelab.
+path "kv/data/k8s/opensandbox-api" {
+  capabilities = ["read"]
+}
+
+path "kv/metadata/k8s/opensandbox-api" {
+  capabilities = ["read"]
+}

infrastructure/proxmox/ansible/roles/vyos_router/defaults/main.yml

@@ -83,6 +83,11 @@ vyos_postgresql_primary_address: "10.60.0.11"
 vyos_sandbox_k3s_api_address: "10.60.0.13"
 vyos_sandbox_k3s_api_interface: eth1
 vyos_sandbox_k3s_api_port: 6443
+vyos_opensandbox_api_port: 8080
+vyos_opensandbox_node_port: 30080
+# Lifecycle create is synchronous and a cold Kata image pull can exceed the
+# HAProxy 50-second default. Keep this below OpenSandbox's overall timeout.
+vyos_opensandbox_api_timeout: 600
 vyos_sandbox_k3s_servers:
   - name: sandbox1
     address: "10.60.0.11"

infrastructure/proxmox/ansible/roles/vyos_router/templates/vyos.conf.j2

@@ -15,6 +15,20 @@ set interfaces ethernet {{ i.iface }} address {{ i.address }}
 set interfaces ethernet {{ i.iface }} description '{{ i.description }}'
 {% endfor %}
 
+{# OpenSandbox stays on a NodePort; VyOS owns the stable routed frontend. #}
+set load-balancing haproxy service opensandbox-api listen-address {{ vyos_sandbox_k3s_api_address }}
+set load-balancing haproxy service opensandbox-api port {{ vyos_opensandbox_api_port }}
+set load-balancing haproxy service opensandbox-api mode 'tcp'
+set load-balancing haproxy service opensandbox-api backend 'opensandbox-api'
+set load-balancing haproxy service opensandbox-api timeout client {{ vyos_opensandbox_api_timeout }}
+set load-balancing haproxy backend opensandbox-api mode 'tcp'
+set load-balancing haproxy backend opensandbox-api timeout server {{ vyos_opensandbox_api_timeout }}
+{% for server in vyos_sandbox_k3s_servers %}
+set load-balancing haproxy backend opensandbox-api server {{ server.name }} address {{ server.address }}
+set load-balancing haproxy backend opensandbox-api server {{ server.name }} port {{ vyos_opensandbox_node_port }}
+set load-balancing haproxy backend opensandbox-api server {{ server.name }} check
+{% endfor %}
+
 {# --- default route out; OSPF carries the rest --- #}
 set protocols static route 0.0.0.0/0 next-hop {{ vyos_lan_gateway }}
 

infrastructure/sandbox-cluster/README.md

@@ -18,7 +18,8 @@ Flux 管理以下 Kubernetes 资源：
 - Kata Containers 和 CI 专用的 `block-plain` RuntimeClass；
 - SPIRE Agent、SPIFFE CSI Driver 与 workload identity 声明；
 - vmagent、kube-state-metrics、kubelet/cAdvisor scrape 配置和告警；
-- OpenSandbox operator/server、`ci-pod` 与 `ci-vm` Pools。
+- sandbox External Secrets Operator、OpenSandbox controller/server；CI Pool 与 runner
+  调度器由 runner 项目接入。
 
 同一个对象只能有一个 owner。Ansible 不直接部署上述集群内 workload；Flux 不管理
 LXC、K3s datastore 或 K3s 本身。
@@ -92,6 +93,13 @@ ansible-playbook site.yml
 ansible-playbook k3s.yml
 ```
 
+只 reconcile Flux controllers 与 root sync manifest（不触碰 LXC、PostgreSQL 或
+K3s lifecycle）：
+
+```bash
+ansible-playbook flux.yml
+```
+
 K3s 外部 datastore URI 由运行时 `SANDBOX_K3S_DB_PASSWORD` 生成，密码在 URI 中
 进行 URL 编码，最终仅持久化于节点 root 可读的 `/etc/rancher/k3s/config.yaml`
 （mode `0600`）。首节点生成的 K3s join token 仅在同一次 Ansible run 内传给第二节点；
@@ -106,7 +114,9 @@ ansible-playbook verify.yml
 当前已经声明 LXC 生命周期、最小 OS baseline、PostgreSQL 和 K3s，包括系统级
 homelab CA trust。Flux `v2.9.5` controllers 与 root sync 也由 Ansible 通过 K3s
 server manifests 管理；root 使用 homelab CA 访问公开 Gitea 仓库，不保存 Git token。
-集群内 workload 由 `clusters/sandbox/` 分阶段纳入 Flux。
+集群内 workload 由 `clusters/sandbox/` 分阶段纳入 Flux。root Kustomization 的健康检查
+timeout 为 40 分钟，用于覆盖 Kata 等首次安装时会逐节点重启 K3s 的子
+Kustomization；各子项仍保留自己的更短 timeout，故障会在对应子项先行暴露。
 
 ## SPIRE 跨集群 bootstrap
 

infrastructure/sandbox-cluster/ansible/flux.yml

@@ -0,0 +1,6 @@
+---
+- name: Reconcile Flux controllers and root sync
+  hosts: sandbox1
+  gather_facts: false
+  roles:
+    - sandbox_flux

infrastructure/sandbox-cluster/ansible/roles/sandbox_flux/templates/flux-sync.yaml.j2

@@ -35,5 +35,5 @@ spec:
   sourceRef:
     kind: GitRepository
     name: flux-system
-  timeout: 3m
+  timeout: 40m
   wait: true

platform/dynamic-runner/README.md

@@ -1,51 +1,107 @@
-# Gitea dynamic runner controller
+# Gitea dynamic runner
 
-此目录只管理 homelab 中的 controller 部署。controller、worker、Cloud Hypervisor
+本目录部署单副本 Go controller，在同一进程运行 RunnerService scheduler、原生
-launcher 和 guest runner 的源码与发布位于独立仓库
+Kubernetes Pod worker、OpenSandbox VM worker 和 SPIFFE mTLS facade：
-`panxiao81/gitea-dynamic-runner`。
-
-当前 bootstrap controller 接收 Gitea `workflow_job` webhook，将 `[self-hosted, pod]` 和
-`[self-hosted, vm]` 的 queued job 分别发布到 NATS。Pod worker 在本集群创建一次性
-privileged host runner；Docker、BuildKit 和 kind 由 workflow 自行 setup。内部
-endpoint：
 
 ```text
-http://dynamic-runner-controller.dynamic-runner.svc.cluster.local:8787/webhook
+Gitea RunnerService -> scheduler -> JetStream ci.runner.pod
+                                      |
+                                      v
+                              homelab Kubernetes Pod
+                                      |
+                         SPIFFE mTLS RunnerService facade
+                                      |
+                                      v
+                                     Gitea
 ```
 
-OpenBao 路径：
+Pod backend 不经过 OpenSandbox。VM backend 后续启用时才访问 VyOS 暴露的
+OpenSandbox Lifecycle API；本目录不修改 sandbox 平台侧 ESO、Bao Terraform 或
+OpenSandbox chart 所有权边界。
 
-- `kv/k8s/nats.ci_producer_password`：已有 NATS producer 密码。
+## Canary 安全边界
-- `kv/k8s/nats.ci_worker_password`：已有 NATS worker 密码。
-- `kv/k8s/dynamic-runner.webhook_secret`：Gitea webhook HMAC secret。
-- `kv/k8s/gitea-runner.token`：现有 instance runner registration token。
 
-首期 controller 与 runner 镜像由 laptop 本机构建后导入 k3s containerd，作为 CI
+- Deployment 为单副本，滚动策略固定 `maxSurge: 1`、`maxUnavailable: 0`，保证新
-发布链路建立前的 bootstrap。部署使用 `imagePullPolicy: Never`。正式发布 workflow
+  facade Ready 后才终止旧实例。scheduler 必须通过 Kubernetes Lease 保持单 leader，
-获得专用 SPIFFE ID 后，必须将 image 改为 zot digest 并移除本地导入步骤。
+  不能依赖 Recreate 避免重复领取。
+- scheduler 使用一个 runner registration，并按总容量启动并发 `FetchTask` goroutine；
+  `POD_CAPACITY=4` 与 `VM_CAPACITY=1` 分别限制两个 durable consumer 和 backend
+  admission pool。池满时 assignment 保持 JetStream pending，任一 backend 不占用
+  另一方的执行槽位，也不会创建超出容量的 workload。
+- rollout 重叠期间只有持有 `Lease/dynamic-runner-scheduler` 的 controller 执行
+  `FetchTask`；所有 Ready 实例都可通过 backend metadata 恢复 claim 并服务 facade。
+- executor 镜像使用 digest；Pod 以 UID 2000 运行，SPIRE `ClusterStaticEntry` 同时绑定
+  具体 Pod UID 与 `unix:uid:2000`。
+- facade 通过仅内网可路由的 `192.168.10.127:30443` NodePort 提供给 sandbox executor；
+  双方使用 Workload API X509-SVID mTLS，并按 SPIFFE ID 而不是 IP/DNS 名验证服务端。
+  该入口不经过公网或 Cloudflare Tunnel。
+- controller 的 SPIFFE ID 固定为
+  `spiffe://ddupan.top/ns/dynamic-runner/sa/dynamic-runner-controller`。
+- NATS 保留既有最小权限分离：`ci-producer` 仅 publish，`ci-worker` 仅 pull/ACK。
 
-## 身份绑定
+facade claim registry 在启动时从 Pod/OpenSandbox metadata 恢复；scheduler leadership
+由 Kubernetes Lease 持久化协调。Deployment 仍保持 `replicas: 1`，滚动更新期间允许
+一个额外 Pod 提供 facade 连续性。
 
-queued webhook 只负责创建没有业务身份的 Pod。runner 实际领取任务后，Gitea 的
+## Secret 边界
-`in_progress` webhook 会携带实际 `runner_name`；controller 将 binding 消息发布到
-NATS，Pod worker 再给对应 Pod 添加：
 
-```text
+`ExternalSecret/dynamic-runner` 从既有 `ClusterSecretStore/openbao` 读取：
-ci.ddupan.top/identity-bound=true
+
-ci.ddupan.top/spiffe-path=<owner>/<repository>/<percent-encoded-job-name>
+- `kv/k8s/nats`：producer/worker 密码；
+- `kv/k8s/opensandbox-api:api_key`：保留给后续 VM worker；
+- `kv/k8s/gitea-runner:token`：保留的 runner registration token；
+- `kv/k8s/dynamic-runner`：scheduler UUID/token、facade HMAC key 和回滚所需 webhook secret。
+
+scheduler credential 由官方 Gitea Runner v3.5.0 一次注册生成；它只挂载到 controller，
+不会进入 executor。facade capability key 至少 32 字节，controller 为每个 assignment
+确定性生成独立 capability。不要打印 Secret、创建静态 Bao token或把 credential 写入
+Git。OpenBao 写入使用本机 SPIFFE JWT 换取的短期 `local-development` token。
+
+## Workflow 身份与依赖配置
+
+workflow 可复用 [`panxiao81/ci-actions`](https://git.ddupan.top/panxiao81/ci-actions)
+中的 `spiffe-openbao-login@v1` 和 `setup-nexus@v1`。这不改变 runner 的权限边界：
+runner 仅提供 Node.js 20、`spire-agent` 与 Workload API socket，workflow 负责声明 Bao
+role、audience 和具体用途，目标服务 policy 决定是否授权。短期 Bao token 会进入
+Actions job 临时文件，因此这些 Action 只允许在本目录管理的一次性 Pod/VM executor
+中使用，不能迁移到共享或持久 runner。
+
+匿名读取 Nexus public repository 只需 `setup-nexus@v1`，不应为了依赖下载额外申请
+Bao 凭据；需要发布制品时再为对应 repository 建立独立 service account 与最小权限
+policy。
+
+## 首次验收
+
+合并后先观察 Flux 与 controller，不要立即开启 VM：
+
+```bash
+flux reconcile kustomization dynamic-runner --with-source
+kubectl -n dynamic-runner wait externalsecret/dynamic-runner \
+  --for=condition=Ready --timeout=2m
+kubectl -n dynamic-runner rollout status deploy/dynamic-runner-controller --timeout=5m
+kubectl -n dynamic-runner logs deploy/dynamic-runner-controller -f
 ```
 
-`ClusterSPIFFEID/gitea-dynamic-runner` 只匹配已经绑定的 Pod，并签发
+确认 scheduler 只领取一条 `[self-hosted,pod]` task，然后验证：
-`spiffe://ddupan.top/ci/<owner>/<repository>/<job-name>`。runner 的 job-start hook 在
-SVID 可用之前不会放行第一步，因此不能根据 queued 事件错配身份。
 
-每个 runner Pod 使用 `gitea-dynamic-runner` ServiceAccount。该 ServiceAccount 没有
+1. assignment message 进入并离开 durable `pod` consumer；
-Kubernetes API 权限；只有 `dynamic-runner-pod-worker` ServiceAccount 能在本 namespace
+2. `gitea-task-<task-id>` Pod 创建，取得实际 Pod UID；
-create/get/patch/delete Pod。
+3. 同名 `ClusterStaticEntry` 的 parent ID 包含该 UID，SPIFFE ID 使用 repository/job key；
+4. 官方 runner v3.5.0 经 facade claim 精确 task，Gitea 实时收到日志和终态；
+5. Pod consumer 达到 capacity 时 VM consumer 仍可独立接受任务。
 
-长期实现将由兼容 Gitea RunnerService 的 scheduler 直接领取 task，再交给 Pod/VM
+Pod 与 VM 都在 Gitea 接受终态后先把 terminal marker 写入各自 backend metadata，再由
-executor；届时删除 webhook、临时 runner 注册和 identity binding 消息。跟踪见
+lifecycle reconciler 删除执行器。首次 VM 测试仍须观察 BatchSandbox、Pod 与
-`panxiao81/gitea-dynamic-runner` issue #7。
+ClusterStaticEntry 全部消失；完整自动清理通过前不得提高 `POD_CAPACITY` 或
+`VM_CAPACITY`。
 
-Gitea webhook 只订阅 `workflow_job`，content type 使用 JSON，secret 与 Bao 中值
+VM backend 已通过 `vm-dev` canary 完成 Docker、kind、SPIFFE 和完整生命周期测试。
-一致。不要启用 `send_everything`，否则 controller 会收到无关仓库事件。
+生产调度标签为 `[self-hosted, vm]`，初始保持 `VM_CAPACITY=1`；扩容前先观察实际任务的
+资源水位、等待时间以及 OpenSandbox 是否存在 terminal sandbox 残留。
+
+## 回滚
+
+若 controller 在领取 task 前失败，回滚到前一 commit 的 Python controller/Pod worker
+manifests。若已经创建 `gitea-task-*` Pod，先保留现场并核对 Gitea task 状态，不能直接
+重启 scheduler 造成重复执行。scheduler registration 和 capability key 保留在 Bao，
+回滚不需要删除或打印它们。

platform/dynamic-runner/clusterspiffeid.yaml

@@ -1,17 +1,16 @@
 apiVersion: spire.spiffe.io/v1alpha1
 kind: ClusterSPIFFEID
 metadata:
-  name: gitea-dynamic-runner
+  name: dynamic-runner-controller
 spec:
   className: spire-mgmt-spire
-  spiffeIDTemplate: 'spiffe://{{ .TrustDomain }}/ci/{{ index .PodMeta.Annotations "ci.ddupan.top/spiffe-path" }}'
+  spiffeIDTemplate: spiffe://ddupan.top/ns/dynamic-runner/sa/dynamic-runner-controller
   namespaceSelector:
     matchLabels:
       kubernetes.io/metadata.name: dynamic-runner
   podSelector:
     matchLabels:
-      app.kubernetes.io/name: gitea-dynamic-runner
+      app.kubernetes.io/name: dynamic-runner-controller
-      ci.ddupan.top/identity-bound: "true"
   workloadSelectorTemplates:
     - k8s:ns:dynamic-runner
-    - k8s:sa:gitea-dynamic-runner
+    - k8s:sa:dynamic-runner-controller

platform/dynamic-runner/deployment.yaml

@@ -5,6 +5,11 @@ metadata:
   namespace: dynamic-runner
 spec:
   replicas: 1
+  strategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxSurge: 1
+      maxUnavailable: 0
   selector:
     matchLabels:
       app.kubernetes.io/name: dynamic-runner-controller
@@ -14,17 +19,10 @@ spec:
         app.kubernetes.io/name: dynamic-runner-controller
     spec:
       serviceAccountName: dynamic-runner-controller
-      automountServiceAccountToken: false
       initContainers:
         - name: fetch-internal-ca
           image: curlimages/curl:8.16.0@sha256:463eaf6072688fe96ac64fa623fe73e1dbe25d8ad6c34404a669ad3ce1f104b6
-          args:
+          args: [--fail, --silent, --show-error, --output, /trust/ca.pem, https://bao.ad.ddupan.top:8200/v1/pki/ca/pem]
-            - --fail
-            - --silent
-            - --show-error
-            - --output
-            - /trust/ca.pem
-            - https://bao.ad.ddupan.top:8200/v1/pki/ca/pem
           securityContext:
             allowPrivilegeEscalation: false
             capabilities:
@@ -40,40 +38,89 @@ spec:
               mountPath: /trust
       containers:
         - name: controller
-          # Bootstrap import on laptop. Replace with a zot digest after the
+          image: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-controller@sha256:7374a08a238ca1c76a9aee33c57520ae0edaa9b5b443708487d5b3a69a61da4e
-          # repository's image publishing workflow has a dedicated identity.
+          imagePullPolicy: IfNotPresent
-          image: gitea-dynamic-runner-controller:0.3.0-bootstrap
+          args: [controller]
-          imagePullPolicy: Never
           env:
+            - name: COMPONENTS
+              value: scheduler,pod-worker,vm-worker
+            - name: GITEA_INSTANCE_URL
+              value: https://git.ddupan.top
+            - name: GITEA_RUNNER_UUID_FILE
+              value: /run/dynamic-runner-secrets/scheduler-uuid
+            - name: GITEA_RUNNER_TOKEN_FILE
+              value: /run/dynamic-runner-secrets/scheduler-token
             - name: NATS_URL
               value: tls://nats.ad.ddupan.top:4222
             - name: NATS_CA_FILE
               value: /run/trust/ca.pem
-            - name: NATS_PASSWORD_FILE
+            - name: NATS_PRODUCER_PASSWORD_FILE
               value: /run/dynamic-runner-secrets/nats-password
-            - name: WEBHOOK_SECRET_FILE
+            - name: NATS_WORKER_PASSWORD_FILE
-              value: /run/dynamic-runner-secrets/webhook-secret
+              value: /run/dynamic-runner-secrets/nats-worker-password
+            - name: RUNNER_FACADE_CAPABILITY_KEY_FILE
+              value: /run/dynamic-runner-secrets/facade-capability-key
+            - name: RUNNER_FACADE_LISTEN
+              value: :8443
+            - name: RUNNER_FACADE_URL
+              value: https://192.168.10.127:30443
+            - name: RUNNER_FACADE_SPIFFE_ID
+              value: spiffe://ddupan.top/ns/dynamic-runner/sa/dynamic-runner-controller
+            - name: SPIFFE_ENDPOINT_SOCKET
+              value: unix:///run/spire/agent-sockets/spire-agent.sock
+            - name: POD_NAMESPACE
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.namespace
+            - name: POD_EXECUTOR_IMAGE
+              value: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-runner@sha256:325d75a208b1a1c6f3b4d4705e47bbc58b34733edcd12f689c60769ab6772a59
+            - name: POD_SERVICE_ACCOUNT
+              value: gitea-dynamic-runner
+            - name: POD_EXECUTOR_UID
+              value: "2000"
+            - name: POD_CAPACITY
+              value: "4"
+            - name: OPENSANDBOX_API
+              value: http://10.60.0.13:8080
+            - name: OPENSANDBOX_API_KEY_FILE
+              value: /run/dynamic-runner-secrets/opensandbox-api-key
+            - name: OPENSANDBOX_POOL
+              value: ci-vm
+            - name: VM_CAPACITY
+              value: "1"
+            - name: VM_RUNNER_LABEL
+              value: vm
+            - name: VM_TIMEOUT_SECONDS
+              value: "14400"
+            - name: SPIRE_CLUSTER
+              value: homelab
+            - name: SPIRE_CLASS
+              value: spire-mgmt-spire
+            - name: SPIRE_AGENT_ID
+              value: spiffe://ddupan.top/spire/agent/k8s_psat/homelab/cd2d0233-c4ea-4031-8327-e7e359e766dd
+            - name: SSL_CERT_FILE
+              value: /run/trust/ca.pem
           ports:
-            - name: http
+            - name: facade
-              containerPort: 8787
+              containerPort: 8443
           readinessProbe:
-            httpGet:
+            httpGet: null
-              path: /healthz
+            tcpSocket:
-              port: http
+              port: facade
             periodSeconds: 5
           livenessProbe:
-            httpGet:
+            httpGet: null
-              path: /healthz
+            tcpSocket:
-              port: http
+              port: facade
-            initialDelaySeconds: 10
+            initialDelaySeconds: 15
             periodSeconds: 10
           resources:
             requests:
-              cpu: 25m
+              cpu: 50m
-              memory: 32Mi
+              memory: 64Mi
             limits:
-              cpu: 250m
+              cpu: 500m
-              memory: 128Mi
+              memory: 256Mi
           securityContext:
             allowPrivilegeEscalation: false
             capabilities:
@@ -91,6 +138,9 @@ spec:
             - name: trust
               mountPath: /run/trust
               readOnly: true
+            - name: spire-agent-socket
+              mountPath: /run/spire/agent-sockets
+              readOnly: true
       securityContext:
         fsGroup: 65532
         fsGroupChangePolicy: OnRootMismatch
@@ -104,3 +154,7 @@ spec:
         - name: trust
           emptyDir:
             sizeLimit: 1Mi
+        - name: spire-agent-socket
+          csi:
+            driver: csi.spiffe.io
+            readOnly: true

platform/dynamic-runner/external-secret.yaml

@@ -20,6 +20,10 @@ spec:
       remoteRef:
         key: k8s/nats
         property: ci_worker_password
+    - secretKey: opensandbox-api-key
+      remoteRef:
+        key: k8s/opensandbox-api
+        property: api_key
     - secretKey: webhook-secret
       remoteRef:
         key: k8s/dynamic-runner
@@ -28,3 +32,15 @@ spec:
       remoteRef:
         key: k8s/gitea-runner
         property: token
+    - secretKey: scheduler-uuid
+      remoteRef:
+        key: k8s/dynamic-runner
+        property: scheduler_uuid
+    - secretKey: scheduler-token
+      remoteRef:
+        key: k8s/dynamic-runner
+        property: scheduler_token
+    - secretKey: facade-capability-key
+      remoteRef:
+        key: k8s/dynamic-runner
+        property: facade_capability_key

platform/dynamic-runner/kustomization.yaml

@@ -6,5 +6,4 @@ resources:
   - rbac.yaml
   - clusterspiffeid.yaml
   - deployment.yaml
-  - pod-worker-deployment.yaml
   - service.yaml

platform/dynamic-runner/pod-worker-deployment.yaml

@@ -1,100 +0,0 @@
-apiVersion: apps/v1
-kind: Deployment
-metadata:
-  name: dynamic-runner-pod-worker
-  namespace: dynamic-runner
-spec:
-  replicas: 1
-  selector:
-    matchLabels:
-      app.kubernetes.io/name: dynamic-runner-pod-worker
-  template:
-    metadata:
-      labels:
-        app.kubernetes.io/name: dynamic-runner-pod-worker
-    spec:
-      serviceAccountName: dynamic-runner-pod-worker
-      initContainers:
-        - name: fetch-internal-ca
-          image: curlimages/curl:8.16.0@sha256:463eaf6072688fe96ac64fa623fe73e1dbe25d8ad6c34404a669ad3ce1f104b6
-          args:
-            - --fail
-            - --silent
-            - --show-error
-            - --output
-            - /trust/ca.pem
-            - https://bao.ad.ddupan.top:8200/v1/pki/ca/pem
-          securityContext:
-            allowPrivilegeEscalation: false
-            capabilities:
-              drop: [ALL]
-            readOnlyRootFilesystem: true
-            runAsNonRoot: true
-            runAsUser: 101
-            runAsGroup: 102
-            seccompProfile:
-              type: RuntimeDefault
-          volumeMounts:
-            - name: trust
-              mountPath: /trust
-      containers:
-        - name: pod-worker
-          image: gitea-dynamic-runner-controller:0.3.0-bootstrap
-          imagePullPolicy: Never
-          command: [/venv/bin/gitea-dynamic-runner-pod-worker]
-          env:
-            - name: NATS_URL
-              value: tls://nats.ad.ddupan.top:4222
-            - name: NATS_CA_FILE
-              value: /run/trust/ca.pem
-            - name: NATS_PASSWORD_FILE
-              value: /run/dynamic-runner-secrets/nats-worker-password
-            - name: RUNNER_NAMESPACE
-              valueFrom:
-                fieldRef:
-                  fieldPath: metadata.namespace
-            - name: RUNNER_IMAGE
-              value: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-runner@sha256:4c61f6315453d68a827ee9542f1345ed86576eaaf62325eb803c8fe3f06ddf2a
-            - name: RUNNER_SERVICE_ACCOUNT
-              value: gitea-dynamic-runner
-            - name: RUNNER_TOKEN_SECRET
-              value: dynamic-runner
-            - name: RUNNER_CAPACITY
-              value: "4"
-          resources:
-            requests:
-              cpu: 25m
-              memory: 32Mi
-            limits:
-              cpu: 250m
-              memory: 128Mi
-          securityContext:
-            allowPrivilegeEscalation: false
-            capabilities:
-              drop: [ALL]
-            readOnlyRootFilesystem: true
-            runAsNonRoot: true
-            runAsUser: 65532
-            runAsGroup: 65532
-            seccompProfile:
-              type: RuntimeDefault
-          volumeMounts:
-            - name: secret
-              mountPath: /run/dynamic-runner-secrets
-              readOnly: true
-            - name: trust
-              mountPath: /run/trust
-              readOnly: true
-      securityContext:
-        fsGroup: 65532
-        fsGroupChangePolicy: OnRootMismatch
-        seccompProfile:
-          type: RuntimeDefault
-      volumes:
-        - name: secret
-          secret:
-            secretName: dynamic-runner
-            defaultMode: 0400
-        - name: trust
-          emptyDir:
-            sizeLimit: 1Mi

platform/dynamic-runner/rbac.yaml

@@ -6,12 +6,6 @@ metadata:
 ---
 apiVersion: v1
 kind: ServiceAccount
-metadata:
-  name: dynamic-runner-pod-worker
-  namespace: dynamic-runner
----
-apiVersion: v1
-kind: ServiceAccount
 metadata:
   name: gitea-dynamic-runner
   namespace: dynamic-runner
@@ -19,23 +13,48 @@ metadata:
 apiVersion: rbac.authorization.k8s.io/v1
 kind: Role
 metadata:
-  name: dynamic-runner-pod-worker
+  name: dynamic-runner-controller
   namespace: dynamic-runner
 rules:
   - apiGroups: [""]
     resources: [pods]
-    verbs: [create, get, patch, delete]
+    verbs: [create, get, list, watch, patch, delete]
+  - apiGroups: [coordination.k8s.io]
+    resources: [leases]
+    verbs: [create, get, list, watch, update, patch]
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: RoleBinding
 metadata:
-  name: dynamic-runner-pod-worker
+  name: dynamic-runner-controller
   namespace: dynamic-runner
 roleRef:
   apiGroup: rbac.authorization.k8s.io
   kind: Role
-  name: dynamic-runner-pod-worker
+  name: dynamic-runner-controller
 subjects:
   - kind: ServiceAccount
-    name: dynamic-runner-pod-worker
+    name: dynamic-runner-controller
+    namespace: dynamic-runner
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: dynamic-runner-spiffe-entries
+rules:
+  - apiGroups: [spire.spiffe.io]
+    resources: [clusterstaticentries]
+    verbs: [create, get, delete]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: dynamic-runner-spiffe-entries
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: dynamic-runner-spiffe-entries
+subjects:
+  - kind: ServiceAccount
+    name: dynamic-runner-controller
     namespace: dynamic-runner

platform/dynamic-runner/service.yaml

@@ -4,9 +4,11 @@ metadata:
   name: dynamic-runner-controller
   namespace: dynamic-runner
 spec:
+  type: NodePort
   selector:
     app.kubernetes.io/name: dynamic-runner-controller
   ports:
-    - name: http
+    - name: facade
-      port: 8787
+      port: 8443
-      targetPort: http
+      targetPort: facade
+      nodePort: 30443

platform/gitea-runner/README.md

@@ -1,114 +0,0 @@
-# Gitea Actions runner
-
-This is the bootstrap runner for Gitea Actions. One persistent runner Pod accepts
-up to four jobs; each job runs in a dynamically created container inside a
-Docker-in-Docker daemon. The official chart runs DinD privileged. Rootless DinD
-would still be privileged and is blocked by the node's AppArmor user-namespace
-policy, so this deployment uses regular DinD instead of weakening that host-wide
-policy. Only trusted workflows may target this runner.
-
-DinD 同时使用 `--mtu=1450` 和
-`--default-network-opt=bridge=com.docker.network.driver.mtu=1450`，与 k3s Pod 的
-`eth0` 一致。前者只覆盖 Docker 默认 bridge；act 为每个 job 创建 user-defined
-bridge，必须由后者设置默认 MTU。不要在未验证节点 Pod MTU 的情况下删除或修改这
-两个参数：MTU 1500 的 job 容器虽然能够解析 GitHub、甚至建立 TCP 连接，但较大的
-TLS 数据包会在嵌套网络路径中丢失，表现为 `github.com` / `api.github.com` 超时或
-`setup-go` 每次请求卡满 6 分钟后重试。Pod 网络和默认 Docker bridge 正常不代表
-Actions job bridge 正常。
-
-The runner is registered at instance scope so it is available to every repository
-on this Gitea instance. Repository permissions and protected-branch review are
-therefore the security boundary; do not enable Actions for untrusted repositories.
-
-The runner registration token is authoritative in OpenBao at
-`kv/k8s/gitea-runner`. External Secrets Operator projects its `token` property to
-the `gitea-runner-token` Secret. Never put the token in this directory or a Helm
-command line.
-
-## SPIRE 与 OCI 发布
-
-runner Pod 使用专用 ServiceAccount `gitea-actions`，并由精确匹配 namespace、
-ServiceAccount 隐含的 Pod、以及 chart labels 的 `ClusterSPIFFEID` 获得：
-
-```text
-spiffe://ddupan.top/ci/gitea-actions
-```
-
-SPIFFE CSI socket 同时只读挂载到 runner 和 DinD。act 的 volume allowlist 只允许
-`/run/spire/agent-sockets`；workflow 仍必须在 job container 中显式请求该 bind
-mount。原因是 bind mount 由 DinD 内的 dockerd 解析，只挂 runner 容器无法让 job
-访问 Workload API。
-
-该身份不是通用 registry 管理员。zot 只对明确列出的 CI 镜像仓库授予
-`read/create/update`，不授予 delete 或其他仓库写入。workflow 应获取
-`aud=zot` 的短期 JWT-SVID，并经 stdin 传给 registry client，不得把 JWT、X.509
-SVID 或 Docker auth 写入 workspace/artifact。
-
-## Flux 接管状态
-
-该 release 最初通过下述 review-first 流程手动 bootstrap。下一个 GitOps 阶段将
-使用 Flux `HelmRelease` 接管它，并首先固定现有 chart `0.1.1`，不在接管 PR 中升级。
-
-迁移前审计发现：Helm 保存的 user-supplied values 和 release manifest 仍描述失败的
-rootless DinD 尝试，但 live StatefulSet 与本目录 `values.yaml` 都已经使用 regular
-DinD。首次 reconcile 的验收条件是修正 Helm 存储状态，同时 live Pod spec、PVC
-identity、runner capacity 和在线状态保持不变。接管稳定后再用独立 PR 升级 chart。
-
-接管分两阶段：第一阶段提交 `suspend: true` 的 HelmRelease、HelmRepository 和由
-`values.yaml` 生成的 ConfigMap。Flux 只登记这些对象，不执行 Helm action。合并后
-检查 HelmRepository Ready，并用固定 chart 重复比较期望清单与 live StatefulSet；
-第二阶段解除 suspend。第一阶段已经确认 source Ready、完整 chart render 与 live
-资源零差异，且登记过程中现有 runner 没有 rollout。失败重试使用
-`RetryOnFailure`，不会用 stored rootless release 做 rollback。
-
-## 历史 review-first bootstrap
-
-这是 Flux 安装前执行过的一次性手动部署流程，保留用于恢复和审计：
-
-1. Merge the reviewed PR.
-2. As a Gitea site administrator, create an instance-scoped runner registration
-   token under **Site Administration → Actions → Runners**.
-3. Store it as the `token` property at `kv/k8s/gitea-runner` without exposing it
-   in shell history:
-
-   ```bash
-   read -rsp 'Runner token: ' runner_token
-   printf '%s' "$runner_token" | bao kv put kv/k8s/gitea-runner token=-
-   unset runner_token
-   ```
-
-4. From the updated `main`, create the namespace and ExternalSecret, then wait
-   for `SecretSynced=True`:
-
-   ```bash
-   KUBECONFIG="$HOME/.kube/config" k3s kubectl apply \
-     -f platform/gitea-runner/namespace.yaml
-   KUBECONFIG="$HOME/.kube/config" k3s kubectl apply \
-     -f platform/gitea-runner/external-secret.yaml
-   KUBECONFIG="$HOME/.kube/config" k3s kubectl wait \
-     --namespace gitea-actions \
-     --for=condition=Ready externalsecret/gitea-runner-token \
-     --timeout=60s
-   ```
-
-5. Install chart `actions` version `0.1.1` from
-   `https://dl.gitea.com/charts/` with this `values.yaml`:
-
-   ```bash
-   helm repo add gitea-charts https://dl.gitea.com/charts/
-   helm repo update gitea-charts
-   helm upgrade --install gitea-actions gitea-charts/actions \
-     --namespace gitea-actions \
-     --version 0.1.1 \
-     --values platform/gitea-runner/values.yaml \
-     --wait --timeout 10m
-   ```
-
-6. Confirm the runner is online, then re-run the queued lint workflow.
-
-Do not deploy from an unmerged feature branch. Do not use `--set` for the token.
-
-The 1 GiB PVC preserves `.runner` identity. Docker image layers are ephemeral;
-the Pod has a 20 GiB ephemeral-storage limit. Terraform apply jobs must use a
-workflow concurrency group because runner capacity does not serialize access to
-a shared state.

platform/gitea-runner/clusterspiffeid.yaml

@@ -1,17 +0,0 @@
-apiVersion: spire.spiffe.io/v1alpha1
-kind: ClusterSPIFFEID
-metadata:
-  name: gitea-actions
-spec:
-  className: spire-mgmt-spire
-  spiffeIDTemplate: spiffe://{{ .TrustDomain }}/ci/gitea-actions
-  namespaceSelector:
-    matchLabels:
-      kubernetes.io/metadata.name: gitea-actions
-  podSelector:
-    matchLabels:
-      app.kubernetes.io/instance: gitea-actions
-      app.kubernetes.io/name: actions-runner
-  workloadSelectorTemplates:
-    - k8s:ns:gitea-actions
-    - k8s:sa:gitea-actions

platform/gitea-runner/helmrepository.yaml

@@ -1,8 +0,0 @@
-apiVersion: source.toolkit.fluxcd.io/v1
-kind: HelmRepository
-metadata:
-  name: gitea-charts
-  namespace: gitea-actions
-spec:
-  interval: 1h
-  url: https://dl.gitea.com/charts/

platform/gitea-runner/kustomization.yaml

@@ -1,18 +0,0 @@
-apiVersion: kustomize.config.k8s.io/v1beta1
-kind: Kustomization
-generatorOptions:
-  disableNameSuffixHash: true
-  labels:
-    reconcile.fluxcd.io/watch: Enabled
-configMapGenerator:
-  - name: gitea-actions-values
-    namespace: gitea-actions
-    files:
-      - values.yaml=values.yaml
-resources:
-  - namespace.yaml
-  - serviceaccount.yaml
-  - clusterspiffeid.yaml
-  - external-secret.yaml
-  - helmrepository.yaml
-  - helmrelease.yaml

platform/gitea-runner/serviceaccount.yaml

@@ -1,6 +0,0 @@
-apiVersion: v1
-kind: ServiceAccount
-metadata:
-  name: gitea-actions
-  namespace: gitea-actions
-automountServiceAccountToken: false

platform/gitea-runner/values.yaml

@@ -1,81 +0,0 @@
-enabled: true
-
-giteaRootURL: http://gitea-http.gitea.svc.cluster.local:3000
-existingSecret: gitea-runner-token
-existingSecretKey: token
-
-statefulset:
-  replicas: 1
-  timezone: Etc/UTC
-  serviceAccountName: gitea-actions
-  extraVolumes:
-    - name: spiffe-workload-api
-      csi:
-        driver: csi.spiffe.io
-        readOnly: true
-  securityContext:
-    fsGroup: 1000
-  # Chart 0.1.1 applies this block to both runner and DinD containers.
-  resources:
-    requests:
-      cpu: 250m
-      memory: 512Mi
-      ephemeral-storage: 2Gi
-    limits:
-      cpu: "4"
-      memory: 6Gi
-      ephemeral-storage: 20Gi
-  persistence:
-    size: 1Gi
-
-  runner:
-    registry: docker.io
-    repository: gitea/runner
-    tag: 2.3.0
-    pullPolicy: IfNotPresent
-    extraVolumeMounts:
-      - name: spiffe-workload-api
-        mountPath: /run/spire/agent-sockets
-        readOnly: true
-    config: |
-      log:
-        level: info
-      runner:
-        file: .runner
-        capacity: 4
-        timeout: 3h
-        shutdown_timeout: 3h
-        labels:
-          - self-hosted:docker://docker.gitea.com/runner-images:ubuntu-latest
-      cache:
-        enabled: false
-      container:
-        require_docker: true
-        docker_timeout: 300s
-        # Workflows must still request this exact bind mount explicitly. The
-        # allowlist prevents arbitrary host paths from reaching job containers.
-        valid_volumes:
-          - /run/spire/agent-sockets
-
-  dind:
-    # The node enforces AppArmor's unprivileged-userns restriction, which blocks
-    # rootlesskit even though this chart must run DinD privileged either way.
-    rootless: false
-    registry: docker.io
-    repository: docker
-    tag: 29.7.1-dind
-    pullPolicy: IfNotPresent
-    # Bind mounts are resolved by dockerd, so the CSI socket must exist in the
-    # DinD container as well as in the runner container.
-    extraVolumeMounts:
-      - name: spiffe-workload-api
-        mountPath: /run/spire/agent-sockets
-        readOnly: true
-    # k3s uses a 1450-byte pod MTU. Without matching it here, nested Actions
-    # networks advertise 1500 and GitHub TLS packets disappear on the outer
-    # overlay path while direct pod traffic remains healthy.
-    extraArgs:
-      - --mtu=1450
-      # --mtu only changes Docker's default bridge. act creates a user-defined
-      # bridge per job, so give every new bridge the same explicit default.
-      - --default-network-opt=bridge=com.docker.network.driver.mtu=1450

platform/sandbox-ci-runners/README.md

@@ -0,0 +1,84 @@
+# OpenSandbox Gitea runner Pools
+
+本目录只部署零预热的 `ci-vm` Pool、sandbox 内的 SPIFFE identity controller，以及仅供
+内网 VyOS 转发的 OpenSandbox NodePort。`ci-vm` 使用 `kata-clh-runtime-rs`，每个 VM
+执行单一任务并在结束后删除。普通 Pod job 直接运行在 homelab，不进入 OpenSandbox。
+
+Pool 中 task-executor 接收 Lifecycle API 下发的进程环境。guest-local SPIRE Agent 用
+Pod-bound PSAT 向中央 SPIRE 注册；identity controller 从 BatchSandbox allocation
+取得真实 Pod UID，再创建精确的 `ClusterStaticEntry`。runner 只有拿到请求中的完整
+repository/task SVID 后才领取一次性 Gitea registration token。
+
+VM 与 Pod backend 使用同一个 `gitea-dynamic-runner-runner` executor 镜像。Pool 不运行
+常驻 `docker:dind` sidecar；需要 Docker 的 workflow 在 privileged executor 内按任务启动
+daemon。Kata VM 中的 workflow 必须先把稀疏 ext4 镜像 loop-mount 到
+`/var/lib/docker`，并完成 cgroup v2 nesting 初始化。
+
+`ci-vm` executor 固定请求并限制为 2 CPU/3GiB。真实 kind canary 表明 1 CPU/约 2GiB
+虽然能最终启动全部 control-plane 容器，但无法在 kubeadm 超时前提供可用的 API server；
+该规格是 nested Kubernetes 任务的容量下限，不是用资源掩盖存储阻塞。
+
+这些 `ClusterStaticEntry` 位于 sandbox 集群，由 central SPIRE Server 内的
+`spire-controller-manager-sandbox` 通过受限 external kubeconfig reconcile。必须在
+`platform/spire/values.yaml` 显式启用 external controller-manager 的
+`reconcile.clusterStaticEntries`（chart 默认关闭）；仅看到 CR 存在但没有 status，不算
+身份链路就绪。
+
+## 部署依赖
+
+- OpenSandbox chart 和 CRD 已 Ready；
+- RuntimeClass `kata-clh-runtime-rs` 已存在；
+- 中央 SPIRE 已发布 `ConfigMap/opensandbox/spire-bundle-pem`；
+- VyOS `10.60.0.13:8080` 转发 sandbox1/2 的 NodePort `30080`；
+- runner/controller 镜像均使用 Zot digest，而不是可变 tag。
+
+本目录不读取 OpenBao，也不修改 OpenSandbox 平台侧 ExternalSecret、ClusterSecretStore
+或 Bao policy。OpenSandbox API key 只存在于平台 server Secret 和 homelab controller
+Secret，两边由各自身份读取同一 Bao 资源。
+
+## 上线验收
+
+先确认 Secret 和 API 认证，命令不得输出 key：
+
+```bash
+kubectl -n dynamic-runner wait externalsecret/dynamic-runner --for=condition=Ready --timeout=2m
+kubectl -n dynamic-runner exec deploy/dynamic-runner-controller -- \
+  wget -qO- http://10.60.0.13:8080/health
+```
+
+然后触发一个 `runs-on: [self-hosted, vm]` 的最小 workflow，并同时观察：
+
+```bash
+kubectl -n dynamic-runner logs deploy/dynamic-runner-controller -f
+kubectl -n opensandbox get pool,batchsandbox,pod -w
+kubectl get clusterstaticentry -l app.kubernetes.io/component=opensandbox-identity -w
+kubectl -n opensandbox logs deploy/opensandbox-identity -f
+```
+
+合格证据必须同时包含：
+
+1. Lifecycle create 成功并选择 `ci-vm`；
+2. 分配 Pod 的 `runtimeClassName` 为 `kata-clh-runtime-rs`；
+3. entry 的 parent ID 包含该 Pod 的 UID，SPIFFE ID 使用 repository/task；
+4. Gitea 显示临时 runner 带 `self-hosted,vm` labels 并完成真实任务；
+5. 任务后 BatchSandbox、Pod、ClusterStaticEntry 和临时 runner 均消失。
+
+清理超时对象时只调用 Lifecycle DELETE，不直接删除 Pool Pod。若 controller 已不可用，
+可从 OpenSandbox API 按 metadata 定位 sandbox 后执行 DELETE；不要绕过 API 伪造状态。
+
+## 故障定位
+
+- `401`：检查 homelab ExternalSecret Ready 和文件挂载，不打印 Secret；
+- `PoolCapacityExhausted`：检查 `ci-vm` 的 `poolMax` 及残留 BatchSandbox；
+- runner 等待 SVID：核对 allocation Pod UID、ClusterStaticEntry parentID、guest Agent 日志；
+- runner 等待 token：核对 `192.168.10.127:8787` 的 sandbox 到 homelab 路由；
+- Docker 任务失败：检查 workflow 的 job-local dockerd 日志及
+  `/var/run/docker.sock`；Pool 不提供共享或常驻 daemon。
+- kind node 的 systemd 报 `Failed to create /init.scope` 或 `Structure needs
+  cleaning`：确认 workflow 启动 dockerd 前完成 cgroup v2 nesting 初始化。否则
+  子容器的私有 cgroup namespace 根会变成 `threaded`，systemd 无法创建 domain
+  cgroup。
+- kind 的 kubeadm 卡在 `CreateContainer`，而 nested containerd goroutine 停在
+  `bbolt` 的 `fdatasync`：不要直接把 Kata shared mount 用作 Docker 数据目录。
+  workflow 必须在 guest 内创建稀疏 ext4 镜像并 loop-mount 到
+  `/var/lib/docker`；该镜像随 Sandbox 删除，不得在节点侧遗留。

platform/sandbox-ci-runners/agent-config.yaml

@@ -0,0 +1,31 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: gitea-ci-spire-agent
+  namespace: opensandbox
+data:
+  agent.conf: |
+    agent {
+      data_dir = "/run/spire/data"
+      log_level = "INFO"
+      server_address = "spire-server.ad.ddupan.top"
+      server_port = "8081"
+      socket_path = "/run/spire/agent-sockets/spire-agent.sock"
+      trust_bundle_path = "/run/spire/bundle/bundle.pem"
+      trust_domain = "ddupan.top"
+    }
+
+    plugins {
+      NodeAttestor "k8s_psat" {
+        plugin_data {
+          cluster = "sandbox-kata"
+          token_path = "/run/spire/tokens/token"
+        }
+      }
+      KeyManager "memory" {
+        plugin_data {}
+      }
+      WorkloadAttestor "unix" {
+        plugin_data {}
+      }
+    }

platform/sandbox-ci-runners/identity-controller.yaml

@@ -0,0 +1,41 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: opensandbox-identity
+  namespace: opensandbox
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: opensandbox-identity
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: opensandbox-identity
+    spec:
+      serviceAccountName: opensandbox-identity
+      containers:
+        - name: controller
+          image: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-controller@sha256:dfbfaf2a7aa5951d1dc4e7f941a8cd05ceb4a860ac1ab23eb8be234861edcb55
+          imagePullPolicy: IfNotPresent
+          command: [/venv/bin/gitea-dynamic-runner-opensandbox-identity]
+          resources:
+            requests:
+              cpu: 10m
+              memory: 32Mi
+            limits:
+              cpu: 100m
+              memory: 96Mi
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop: [ALL]
+            readOnlyRootFilesystem: true
+            runAsNonRoot: true
+            runAsUser: 65532
+            runAsGroup: 65532
+            seccompProfile:
+              type: RuntimeDefault
+      securityContext:
+        seccompProfile:
+          type: RuntimeDefault

platform/sandbox-ci-runners/kustomization.yaml

@@ -0,0 +1,22 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - service.yaml
+  - agent-config.yaml
+  - rbac.yaml
+  - identity-controller.yaml
+  - pools.yaml
+patches:
+  # Pod jobs run directly in the homelab cluster. Keep OpenSandbox VM-only.
+  - target:
+      group: sandbox.opensandbox.io
+      version: v1alpha1
+      kind: Pool
+      name: ci-pod
+    patch: |-
+      apiVersion: sandbox.opensandbox.io/v1alpha1
+      kind: Pool
+      metadata:
+        name: ci-pod
+        namespace: opensandbox
+      $patch: delete

platform/sandbox-ci-runners/pools.yaml

@@ -0,0 +1,293 @@
+---
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: ci-vm
+  namespace: opensandbox
+  labels:
+    app.kubernetes.io/name: gitea-ci-vm
+    app.kubernetes.io/component: runner-pool
+spec:
+  capacitySpec:
+    bufferMax: 0
+    bufferMin: 0
+    poolMax: 2
+    poolMin: 0
+  recycleStrategy:
+    type: Delete
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: gitea-ci-vm
+        ci.ddupan.top/backend: vm
+    spec:
+      runtimeClassName: kata-clh-runtime-rs
+      serviceAccountName: gitea-ci
+      restartPolicy: Never
+      terminationGracePeriodSeconds: 30
+      shareProcessNamespace: true
+      securityContext:
+        fsGroup: 2000
+        fsGroupChangePolicy: OnRootMismatch
+      initContainers:
+        - name: task-executor-installer
+          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/task-executor:v0.1.0
+          command: [/bin/sh, -c]
+          args:
+            - cp /workspace/server /opt/opensandbox/task-executor && chmod 0755 /opt/opensandbox/task-executor
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+        - name: execd-installer
+          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.22
+          command: [/bin/sh, -c]
+          args:
+            - cp ./execd /opt/opensandbox/execd && cp ./bootstrap.sh /opt/opensandbox/bootstrap.sh && chmod 0755 /opt/opensandbox/execd /opt/opensandbox/bootstrap.sh
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+      containers:
+        - name: sandbox
+          image: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-runner@sha256:325d75a208b1a1c6f3b4d4705e47bbc58b34733edcd12f689c60769ab6772a59
+          imagePullPolicy: IfNotPresent
+          command: [/opt/opensandbox/task-executor]
+          args:
+            - -listen-addr=0.0.0.0:5758
+            - -log-dir=/tmp
+            - -data-dir=/tmp/tasks
+          env:
+            - name: SANDBOX_MAIN_CONTAINER
+              value: sandbox
+            - name: EXECD_ENVS
+              value: /opt/opensandbox/.env
+            - name: EXECD
+              value: /opt/opensandbox/execd
+            - name: GITEA_INSTANCE_URL
+              value: https://git.ddupan.top
+            - name: HOME
+              value: /data
+          ports:
+            - name: task-executor
+              containerPort: 5758
+          securityContext:
+            privileged: true
+            runAsNonRoot: true
+            runAsUser: 2000
+            runAsGroup: 2000
+          resources:
+            requests:
+              cpu: "2"
+              memory: 3Gi
+            limits:
+              cpu: "2"
+              memory: 3Gi
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+            - name: runner-data
+              mountPath: /data
+            - name: runner-workspace
+              mountPath: /workspace
+            - name: spire-socket
+              mountPath: /run/spire/agent-sockets
+        - name: spire-agent
+          image: ghcr.io/spiffe/spire-agent:1.15.3@sha256:41b0dcd8b258a69db9e2768292a060766fb76fd866e4bc925849981ea1b825ff
+          args: [-config, /run/spire/config/agent.conf]
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop: [ALL]
+          volumeMounts:
+            - name: spire-config
+              mountPath: /run/spire/config
+              readOnly: true
+            - name: spire-bundle
+              mountPath: /run/spire/bundle
+              readOnly: true
+            - name: spire-token
+              mountPath: /run/spire/tokens
+              readOnly: true
+            - name: spire-data
+              mountPath: /run/spire/data
+            - name: spire-socket
+              mountPath: /run/spire/agent-sockets
+      volumes:
+        - name: opensandbox-bin
+          emptyDir: {}
+        - name: runner-data
+          emptyDir: {}
+        - name: runner-workspace
+          emptyDir: {}
+        - name: spire-data
+          emptyDir:
+            medium: Memory
+        - name: spire-socket
+          emptyDir:
+            medium: Memory
+        - name: spire-config
+          configMap:
+            name: gitea-ci-spire-agent
+        - name: spire-bundle
+          configMap:
+            name: spire-bundle-pem
+        - name: spire-token
+          projected:
+            sources:
+              - serviceAccountToken:
+                  audience: spire-server
+                  expirationSeconds: 3600
+                  path: token
+---
+apiVersion: sandbox.opensandbox.io/v1alpha1
+kind: Pool
+metadata:
+  name: ci-pod
+  namespace: opensandbox
+  labels:
+    app.kubernetes.io/name: gitea-ci-pod
+    app.kubernetes.io/component: runner-pool
+spec:
+  capacitySpec:
+    bufferMax: 0
+    bufferMin: 0
+    poolMax: 4
+    poolMin: 0
+  recycleStrategy:
+    type: Delete
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: gitea-ci-pod
+        ci.ddupan.top/backend: pod
+    spec:
+      serviceAccountName: gitea-ci
+      restartPolicy: Never
+      terminationGracePeriodSeconds: 30
+      shareProcessNamespace: true
+      securityContext:
+        fsGroup: 2000
+        fsGroupChangePolicy: OnRootMismatch
+      initContainers:
+        - name: task-executor-installer
+          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/task-executor:v0.1.0
+          command: [/bin/sh, -c]
+          args:
+            - cp /workspace/server /opt/opensandbox/task-executor && chmod 0755 /opt/opensandbox/task-executor
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+        - name: execd-installer
+          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.22
+          command: [/bin/sh, -c]
+          args:
+            - cp ./execd /opt/opensandbox/execd && cp ./bootstrap.sh /opt/opensandbox/bootstrap.sh && chmod 0755 /opt/opensandbox/execd /opt/opensandbox/bootstrap.sh
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+      containers:
+        - name: sandbox
+          image: zot.ad.ddupan.top/panxiao81/gitea-dynamic-runner-runner@sha256:325d75a208b1a1c6f3b4d4705e47bbc58b34733edcd12f689c60769ab6772a59
+          imagePullPolicy: IfNotPresent
+          command: [/opt/opensandbox/task-executor]
+          args:
+            - -listen-addr=0.0.0.0:5758
+            - -log-dir=/tmp
+            - -data-dir=/tmp/tasks
+          env:
+            - name: SANDBOX_MAIN_CONTAINER
+              value: sandbox
+            - name: EXECD_ENVS
+              value: /opt/opensandbox/.env
+            - name: EXECD
+              value: /opt/opensandbox/execd
+            - name: GITEA_INSTANCE_URL
+              value: https://git.ddupan.top
+            - name: HOME
+              value: /data
+            - name: DOCKER_HOST
+              value: unix:///run/docker/docker.sock
+          ports:
+            - name: task-executor
+              containerPort: 5758
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop: [ALL]
+            runAsNonRoot: true
+            runAsUser: 2000
+            runAsGroup: 2000
+          volumeMounts:
+            - name: opensandbox-bin
+              mountPath: /opt/opensandbox
+            - name: runner-data
+              mountPath: /data
+            - name: runner-workspace
+              mountPath: /workspace
+            - name: docker-socket
+              mountPath: /run/docker
+            - name: spire-socket
+              mountPath: /run/spire/agent-sockets
+        - name: spire-agent
+          image: ghcr.io/spiffe/spire-agent:1.15.3@sha256:41b0dcd8b258a69db9e2768292a060766fb76fd866e4bc925849981ea1b825ff
+          args: [-config, /run/spire/config/agent.conf]
+          securityContext:
+            allowPrivilegeEscalation: false
+            capabilities:
+              drop: [ALL]
+          volumeMounts:
+            - name: spire-config
+              mountPath: /run/spire/config
+              readOnly: true
+            - name: spire-bundle
+              mountPath: /run/spire/bundle
+              readOnly: true
+            - name: spire-token
+              mountPath: /run/spire/tokens
+              readOnly: true
+            - name: spire-data
+              mountPath: /run/spire/data
+            - name: spire-socket
+              mountPath: /run/spire/agent-sockets
+        - name: docker
+          image: docker.io/library/docker:29.1.5-dind
+          command: [/bin/sh, -c]
+          args:
+            - test -e /dev/kmsg || mknod /dev/kmsg c 1 11; exec dockerd --host=unix:///run/docker/docker.sock --group=2000 --storage-driver=overlay2
+          securityContext:
+            privileged: true
+          volumeMounts:
+            - name: docker-socket
+              mountPath: /run/docker
+            - name: docker-data
+              mountPath: /var/lib/docker
+      volumes:
+        - name: opensandbox-bin
+          emptyDir: {}
+        - name: runner-data
+          emptyDir: {}
+        - name: runner-workspace
+          emptyDir: {}
+        - name: docker-data
+          emptyDir: {}
+        - name: docker-socket
+          emptyDir:
+            medium: Memory
+        - name: spire-data
+          emptyDir:
+            medium: Memory
+        - name: spire-socket
+          emptyDir:
+            medium: Memory
+        - name: spire-config
+          configMap:
+            name: gitea-ci-spire-agent
+        - name: spire-bundle
+          configMap:
+            name: spire-bundle-pem
+        - name: spire-token
+          projected:
+            sources:
+              - serviceAccountToken:
+                  audience: spire-server
+                  expirationSeconds: 3600
+                  path: token

platform/sandbox-ci-runners/rbac.yaml

@@ -0,0 +1,87 @@
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: gitea-ci
+  namespace: opensandbox
+---
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: opensandbox-identity
+  namespace: opensandbox
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: opensandbox-identity
+  namespace: opensandbox
+rules:
+  - apiGroups: [""]
+    resources: [pods]
+    verbs: [get, list, watch]
+  - apiGroups: [sandbox.opensandbox.io]
+    resources: [batchsandboxes]
+    verbs: [get, list, watch]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: opensandbox-identity
+  namespace: opensandbox
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: opensandbox-identity
+subjects:
+  - kind: ServiceAccount
+    name: opensandbox-identity
+    namespace: opensandbox
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: opensandbox-identity
+rules:
+  - apiGroups: [spire.spiffe.io]
+    resources: [clusterstaticentries]
+    verbs: [create, delete, get, list, watch]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: opensandbox-identity
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: opensandbox-identity
+subjects:
+  - kind: ServiceAccount
+    name: opensandbox-identity
+    namespace: opensandbox
+---
+# The central SPIRE external controller uses this existing sandbox credential
+# to publish a PEM bundle for guest-local Agents. It cannot read Secrets here.
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: spire-runner-bundle-publisher
+  namespace: opensandbox
+rules:
+  - apiGroups: [""]
+    resources: [configmaps]
+    verbs: [create, delete, get, list, patch, update, watch]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: spire-runner-bundle-publisher
+  namespace: opensandbox
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: spire-runner-bundle-publisher
+subjects:
+  - kind: ServiceAccount
+    name: spire-controller-manager
+    namespace: spire-system

platform/sandbox-ci-runners/service.yaml

@@ -0,0 +1,15 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: opensandbox-server-internal
+  namespace: opensandbox-system
+spec:
+  type: NodePort
+  selector:
+    app.kubernetes.io/instance: opensandbox
+    app.kubernetes.io/name: opensandbox-server
+  ports:
+    - name: http
+      port: 80
+      targetPort: http
+      nodePort: 30080

platform/sandbox-external-secrets/README.md

@@ -0,0 +1,56 @@
+# Sandbox External Secrets Operator
+
+本目录在 sandbox 集群部署独立的 External Secrets Operator `2.8.0`，并通过
+`ClusterSecretStore/openbao` 读取 OpenBao KV v2 中共享的 OpenSandbox API key。它不复用
+homelab 集群的 ESO Pod、ServiceAccount 或 Kubernetes auth backend。
+
+## 当前状态
+
+- OpenBao `auth/kubernetes-sandbox`、backend config、`external-secrets` role 与
+  `sandbox-external-secrets` policy 已于 2026-09-18 由 Terraform 创建；apply 后 plan
+  为 zero-diff；
+- `kv/k8s/opensandbox-api` 已由本机 `spiffe://ddupan.top/dev/panxiao81` 身份生成并写入，
+  值未输出或落盘；
+- sandbox ESO operator、`ClusterSecretStore/openbao` 与 OpenSandbox `ExternalSecret`
+  由 Flux 管理；
+- 线上 `ClusterSecretStore/openbao` 为 `Valid/Ready`，`ExternalSecret/opensandbox-api-key`
+  为 `SecretSynced/Ready`；
+- OpenSandbox 已切换到 API key：无 key 请求返回 `401`，正确 key 请求返回 `200`；
+- homelab runner 对同一 key 的投影不在本目录，留给 runner 项目管理。
+
+OpenBao 的 `auth/kubernetes-sandbox`、对应 role、policy、sandbox API 地址与公开 CA
+完全由 `infrastructure/openbao/terraform/` 管理。CA 文件提交到 Git 是刻意设计：CA
+是公开信任材料，版本化后集群重建造成的 trust root 变化会产生可审计的 Terraform diff。
+
+ESO 使用 TokenRequest 生成短期 ServiceAccount JWT。OpenBao 未配置长期
+`token_reviewer_jwt`，而是使用登录 JWT 调用 sandbox TokenReview；因此
+`external-secrets` ServiceAccount 仅额外绑定内建 `system:auth-delegator`。
+
+## 重建顺序
+
+1. 在 OpenBao 写入 OpenSandbox API key：
+
+   ```bash
+   openssl rand -hex 32 | bao kv put kv/k8s/opensandbox-api api_key=-
+   ```
+
+2. 在 `infrastructure/openbao/terraform` 执行 `terraform plan` 和 `terraform apply`，
+   创建 `kubernetes-sandbox` auth mount、backend config、role 与只允许读取
+   `kv/k8s/opensandbox-api` 的最小权限 policy。
+3. 合并 Flux 变更。依次等待 `flux-system/external-secrets-operator`、
+   `flux-system/external-secrets` Ready，再等待 `flux-system/opensandbox` 滚动完成。
+
+operator 与配置拆成两个 Flux Kustomization，确保全新集群先安装 CRD，再声明
+`ClusterSecretStore`；不要为了减少目录而把两层重新合并。
+
+## 验收
+
+```bash
+kubectl get clustersecretstore openbao
+kubectl -n opensandbox-system get externalsecret opensandbox-api-key
+kubectl -n opensandbox-system get secret opensandbox-api-key
+```
+
+只检查 Secret 是否存在及 key 名，不输出 `data`。`ClusterSecretStore` 或
+`ExternalSecret` 不 Ready 时，先检查 `auth/kubernetes-sandbox`，不要临时创建静态
+Bao token Secret。

platform/sandbox-external-secrets/config/auth-delegator.yaml

@@ -0,0 +1,13 @@
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: sandbox-external-secrets-token-review
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: system:auth-delegator
+subjects:
+  - kind: ServiceAccount
+    name: external-secrets
+    namespace: external-secrets

platform/sandbox-external-secrets/config/clustersecretstore.yaml

@@ -0,0 +1,18 @@
+---
+apiVersion: external-secrets.io/v1
+kind: ClusterSecretStore
+metadata:
+  name: openbao
+spec:
+  provider:
+    vault:
+      server: https://bao.ad.ddupan.top:8200
+      path: kv
+      version: v2
+      auth:
+        kubernetes:
+          mountPath: kubernetes-sandbox
+          role: external-secrets
+          serviceAccountRef:
+            name: external-secrets
+            namespace: external-secrets

platform/sandbox-external-secrets/config/kustomization.yaml

@@ -1,4 +1,6 @@
+---
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
-  - pools.yaml
+  - auth-delegator.yaml
+  - clustersecretstore.yaml

platform/sandbox-external-secrets/operator/helmrelease.yaml

@@ -1,31 +1,35 @@
+---
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
-  name: gitea-actions
+  name: external-secrets
-  namespace: gitea-actions
+  namespace: external-secrets
 spec:
   chart:
     spec:
-      chart: actions
+      chart: external-secrets
       interval: 1h
       sourceRef:
         kind: HelmRepository
-        name: gitea-charts
+        name: external-secrets
-      version: 0.1.1
+      version: 2.8.0
   driftDetection:
     mode: enabled
   install:
+    crds: CreateReplace
     strategy:
       name: RetryOnFailure
       retryInterval: 5m
   interval: 30m
-  releaseName: gitea-actions
+  releaseName: external-secrets
-  targetNamespace: gitea-actions
+  targetNamespace: external-secrets
   timeout: 10m
   upgrade:
+    crds: CreateReplace
     strategy:
       name: RetryOnFailure
       retryInterval: 5m
   valuesFrom:
     - kind: ConfigMap
-      name: gitea-actions-values
+      name: external-secrets-values
+      valuesKey: values.yaml

platform/sandbox-external-secrets/operator/kustomization.yaml

@@ -0,0 +1,8 @@
+---
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - namespace.yaml
+  - repository.yaml
+  - values.yaml
+  - helmrelease.yaml

platform/sandbox-external-secrets/operator/namespace.yaml

@@ -1,4 +1,5 @@
+---
 apiVersion: v1
 kind: Namespace
 metadata:
-  name: gitea-actions
+  name: external-secrets

platform/sandbox-external-secrets/operator/repository.yaml

@@ -0,0 +1,9 @@
+---
+apiVersion: source.toolkit.fluxcd.io/v1
+kind: HelmRepository
+metadata:
+  name: external-secrets
+  namespace: external-secrets
+spec:
+  interval: 1h
+  url: https://charts.external-secrets.io

platform/sandbox-external-secrets/operator/values.yaml

@@ -0,0 +1,31 @@
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: external-secrets-values
+  namespace: external-secrets
+data:
+  values.yaml: |
+    replicaCount: 1
+
+    webhook:
+      replicaCount: 1
+      resources:
+        requests: {cpu: 10m, memory: 32Mi}
+        limits: {memory: 128Mi}
+
+    certController:
+      replicaCount: 1
+      resources:
+        requests: {cpu: 10m, memory: 32Mi}
+        limits: {memory: 128Mi}
+
+    resources:
+      requests: {cpu: 10m, memory: 64Mi}
+      limits: {memory: 256Mi}
+
+    serviceAccount:
+      create: true
+      name: external-secrets
+
+    installCRDs: true

platform/sandbox-kata/README.md

@@ -28,6 +28,32 @@ Pod 删除后的 backing-file/VMM 回收和真实构建基准必须作为上线
 由 `VMPodScrape` 写入中央 VictoriaMetrics。它不拥有 Kubernetes API 凭据或 host 写
 权限。
 
+## Guest 内 SPIFFE 身份
+
+Kata guest 不能直接使用 node SPIRE Agent 的 CSI socket。Unix socket 的路径即使通过
+virtio-fs 出现在 guest 中，连接也不能跨 VM 边界。CI Pod 应以 native sidecar 在同一
+guest 内启动临时 SPIRE Agent，并满足以下约束：
+
+1. 外层 Pod 挂载 `audience=spire-server` 的 Pod-bound projected ServiceAccount token；
+2. 内层 Agent 使用 `k8s_psat` 向中央 Server attestation，Server cluster profile 必须
+   启用 `use_pod_uid_for_agent_id`，使 Agent ID 包含 Pod UID；
+3. 调度器为该具体 Agent 创建 registration entry；SPIFFE ID 使用仓库和任务名等稳定的
+   业务语义，Pod UID 只作为 parent binding，不进入业务身份；
+4. Agent 与 workload 通过 guest 内 `emptyDir` 上的 Unix socket 通信；Pod 必须设置
+   `shareProcessNamespace: true`，否则 `unix` workload attestor 无法从共享 `/proc`
+   解析客户端的 `SO_PEERCRED` PID；
+5. 调度器必须在 registration entry 已同步到 Agent 后才放行 workload。Pod 删除后同步
+   删除 entry，并 evict 对应的临时 Agent 记录。
+
+2026-09-17 的 live PoC 已验证：以 Pod UID 作为 parent 的临时 Agent 成功注册，
+`unix:uid:2000` workload 从 guest-local socket 获得
+`spiffe://ddupan.top/ci/poc/task/build` X.509-SVID，Pod 正常退出。PoC 资源随后全部清理，
+中央 SPIRE 恢复声明式配置。
+
+SPIRE 1.15.3 已支持 `use_pod_uid_for_agent_id`，但当前使用的 hardened chart 0.30.2
+尚未把它暴露到 values/template。正式部署应先给 chart 补齐该字段并向上游提交，随后
+采用包含修复的 release；不得把手工修改 Server ConfigMap 作为运行方案。
+
 ## 上线验收
 
 Flux reconciliation 完成后至少确认：
@@ -35,8 +61,8 @@ Flux reconciliation 完成后至少确认：
 1. 两个节点重新回到 Ready，`RuntimeClass/kata-clh-runtime-rs` 存在；
 2. Kata Pod 内核与 LXC host 内核不同，且 `/dev/kvm` 可用；
 3. Cloud Hypervisor API 的 `vm.info.config.memory.shared` 为 `true`；
-4. `spire-smoke` ServiceAccount 的 Kata Pod 可获得
+4. Kata Pod 内层 Agent 的 ID 包含该 Pod UID；只有调度器创建的业务 entry 所匹配的
-   `spiffe://ddupan.top/sandbox/smoke`，错误 ServiceAccount 无法获得身份；
+   workload UID 能从 guest-local socket 获得预期 SPIFFE ID；
 5. block-backed `emptyDir` 上 Docker 使用 `overlay2`，BuildKit 与 kind smoke test
    通过；kind 的 dockerd bootstrap 需要先在 guest 内执行
    `mknod /dev/kmsg c 1 11`；

platform/sandbox-opensandbox-pools/pools.yaml

@@ -1,123 +0,0 @@
----
-apiVersion: sandbox.opensandbox.io/v1alpha1
-kind: Pool
-metadata:
-  name: ci-pod
-  namespace: opensandbox
-spec:
-  template:
-    metadata:
-      labels:
-        ci.ddupan.top/backend: pod
-    spec:
-      containers:
-        - name: sandbox
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/code-interpreter:v1.1.0
-          command: [/opt/opensandbox/task-executor]
-          args: [-listen-addr=0.0.0.0:5758, -log-dir=/tmp]
-          env:
-            - name: SANDBOX_MAIN_CONTAINER
-              value: sandbox
-            - name: EXECD_ENVS
-              value: /opt/opensandbox/.env
-            - name: EXECD
-              value: /opt/opensandbox/execd
-          resources:
-            requests:
-              cpu: 100m
-              memory: 256Mi
-            limits:
-              cpu: "2"
-              memory: 4Gi
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-            - name: sandbox-storage
-              mountPath: /var/lib/sandbox
-      initContainers:
-        - name: task-executor-installer
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/task-executor:v0.1.0
-          command: [/bin/sh, -c]
-          args: [cp /workspace/server /opt/opensandbox/task-executor && chmod 0755 /opt/opensandbox/task-executor]
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-        - name: execd-installer
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.22
-          command: [/bin/sh, -c]
-          args: [cp ./execd /opt/opensandbox/execd && cp ./bootstrap.sh /opt/opensandbox/bootstrap.sh && chmod 0755 /opt/opensandbox/execd /opt/opensandbox/bootstrap.sh]
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-      volumes:
-        - name: opensandbox-bin
-          emptyDir: {}
-        - name: sandbox-storage
-          emptyDir: {}
-  capacitySpec:
-    bufferMax: 1
-    bufferMin: 0
-    poolMax: 4
-    poolMin: 0
----
-apiVersion: sandbox.opensandbox.io/v1alpha1
-kind: Pool
-metadata:
-  name: ci-vm
-  namespace: opensandbox
-spec:
-  template:
-    metadata:
-      labels:
-        ci.ddupan.top/backend: vm
-    spec:
-      runtimeClassName: kata-clh-runtime-rs
-      containers:
-        - name: sandbox
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/code-interpreter:v1.1.0
-          command: [/opt/opensandbox/task-executor]
-          args: [-listen-addr=0.0.0.0:5758, -log-dir=/tmp]
-          env:
-            - name: SANDBOX_MAIN_CONTAINER
-              value: sandbox
-            - name: EXECD_ENVS
-              value: /opt/opensandbox/.env
-            - name: EXECD
-              value: /opt/opensandbox/execd
-          resources:
-            requests:
-              cpu: 250m
-              memory: 512Mi
-            limits:
-              cpu: "4"
-              memory: 8Gi
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-            - name: sandbox-storage
-              mountPath: /var/lib/sandbox
-      initContainers:
-        - name: task-executor-installer
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/task-executor:v0.1.0
-          command: [/bin/sh, -c]
-          args: [cp /workspace/server /opt/opensandbox/task-executor && chmod 0755 /opt/opensandbox/task-executor]
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-        - name: execd-installer
-          image: sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.22
-          command: [/bin/sh, -c]
-          args: [cp ./execd /opt/opensandbox/execd && cp ./bootstrap.sh /opt/opensandbox/bootstrap.sh && chmod 0755 /opt/opensandbox/execd /opt/opensandbox/bootstrap.sh]
-          volumeMounts:
-            - name: opensandbox-bin
-              mountPath: /opt/opensandbox
-      volumes:
-        - name: opensandbox-bin
-          emptyDir: {}
-        - name: sandbox-storage
-          emptyDir: {}
-  capacitySpec:
-    bufferMax: 1
-    bufferMin: 0
-    poolMax: 2
-    poolMin: 0

platform/sandbox-opensandbox/README.md

@@ -1,19 +1,44 @@
-# OpenSandbox
+# Sandbox OpenSandbox
 
-Flux installs the upstream all-in-one OpenSandbox chart pinned to
+本目录在独立 sandbox k3s 集群部署 OpenSandbox controller、server 与 CRD。Flux 从
-`helm/opensandbox/0.2.2` (`8f01e935`). The API is cluster-internal and intentionally runs a
+上游 commit `8f01e935c2cabba778cf37a152033fae062fa0f4` 构建官方 umbrella chart
-single replica until shared server state and HA behaviour have been validated.
+`0.2.2`；该源码渲染结果已与 release `opensandbox-0.2.2.tgz` 对比一致。不要改为跟随
+浮动 branch 或 tag。
 
-`ci-pod` uses `runc`; `ci-vm` uses the separately managed
+server 只提供集群内 `opensandbox-server.opensandbox-system.svc:80` ClusterIP，不部署
-`kata-clh-runtime-rs` RuntimeClass. Both Pools start at zero and create capacity
+Gateway、Ingress 或 LoadBalancer。sandbox workload 位于 `opensandbox` namespace，
-on demand. They currently use the upstream interpreter image to validate the
+默认使用 `kata-clh-runtime-rs`；CI Pool、runner 镜像、动态 SPIFFE registration 均由
-Lifecycle API and Pool allocation independently of the CI scheduler cutover.
+runner 项目后续声明，本目录不预制。
 
-The dynamic runner worker, runner image, guest-local SPIRE Agent and Docker
+## API 认证
-sidecar are introduced only after this layer is Ready. In particular, do not
-mount the host SPIFFE CSI socket into `ci-vm`: Unix sockets do not cross the
-Kata VM boundary.
 
-Smoke test both backends through the same API by creating sandboxes with
+`ExternalSecret/opensandbox-api-key` 从 OpenBao
-`extensions.poolRef` set to `ci-pod` and `ci-vm`, then confirm their
+`kv/k8s/opensandbox-api:api_key` 投影同名 Secret。这个路径不归属于某个 Kubernetes
-BatchSandboxes, Pods and VMMs disappear after deletion.
+集群：sandbox server 与 homelab runner 调度器分别通过自己的 Bao 身份读取。server 只通过
+`secretKeyRef` 读取：
+
+```yaml
+- name: OPENSANDBOX_SERVER_API_KEY
+  valueFrom:
+    secretKeyRef:
+      name: opensandbox-api-key
+      key: api-key
+```
+
+仓库与 Helm values 均不保存 API key。OpenSandbox 不支持更丰富的原生 workload
+authentication；runner 后续读取同一 Bao 路径并在请求头中使用 API key。
+`opensandbox-values` 带 Flux watch label，values 变化会立即触发 Helm reconcile，不依赖
+30 分钟的 HelmRelease interval。
+
+## 验收
+
+合并后等待 `flux-system/opensandbox` 与 `opensandbox-system/opensandbox` Ready，并确认：
+
+```bash
+kubectl get crd batchsandboxes.sandbox.opensandbox.io pools.sandbox.opensandbox.io
+kubectl -n opensandbox-system get deploy,pod,svc
+kubectl get runtimeclass kata-clh-runtime-rs
+```
+
+控制面上线不创建 CI Pool，也不产生 sandbox workload。首个 runner 集成应另行提交 Pool
+与完整的 Lifecycle API smoke test。

platform/sandbox-opensandbox/external-secret.yaml

@@ -1,18 +1,19 @@
+---
 apiVersion: external-secrets.io/v1
 kind: ExternalSecret
 metadata:
-  name: gitea-runner-token
+  name: opensandbox-api-key
-  namespace: gitea-actions
+  namespace: opensandbox-system
 spec:
   refreshInterval: 1h
   secretStoreRef:
     kind: ClusterSecretStore
     name: openbao
   target:
+    name: opensandbox-api-key
     creationPolicy: Owner
-    name: gitea-runner-token
   data:
-    - secretKey: token
+    - secretKey: api-key
       remoteRef:
-        key: k8s/gitea-runner
+        key: k8s/opensandbox-api
-        property: token
+        property: api_key

platform/sandbox-opensandbox/helmrelease.yaml

@@ -1,3 +1,4 @@
+---
 apiVersion: helm.toolkit.fluxcd.io/v2
 kind: HelmRelease
 metadata:
@@ -15,6 +16,7 @@ spec:
   driftDetection:
     mode: enabled
   install:
+    crds: CreateReplace
     strategy:
       name: RetryOnFailure
       retryInterval: 5m
@@ -23,9 +25,11 @@ spec:
   targetNamespace: opensandbox-system
   timeout: 15m
   upgrade:
+    crds: CreateReplace
     strategy:
       name: RetryOnFailure
       retryInterval: 5m
   valuesFrom:
     - kind: ConfigMap
       name: opensandbox-values
+      valuesKey: values.yaml

platform/sandbox-opensandbox/kustomization.yaml

@@ -1,7 +1,11 @@
+---
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization
 resources:
-  - namespaces.yaml
+  - namespace.yaml
+  - external-secret.yaml
   - repository.yaml
+  - template.yaml
   - values.yaml
   - helmrelease.yaml
+  - monitor-scrape.yaml

platform/sandbox-opensandbox/monitor-scrape.yaml

@@ -0,0 +1,16 @@
+---
+apiVersion: operator.victoriametrics.com/v1beta1
+kind: VMPodScrape
+metadata:
+  name: opensandbox-controller
+  namespace: monitoring
+spec:
+  namespaceSelector:
+    matchNames:
+      - opensandbox-system
+  podMetricsEndpoints:
+    - interval: 30s
+      port: metrics
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: opensandbox

platform/sandbox-opensandbox/namespace.yaml

@@ -0,0 +1,14 @@
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: opensandbox-system
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: opensandbox
+  labels:
+    pod-security.kubernetes.io/enforce: privileged
+    pod-security.kubernetes.io/audit: restricted
+    pod-security.kubernetes.io/warn: restricted

platform/sandbox-opensandbox/namespaces.yaml

@@ -1,10 +0,0 @@
----
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: opensandbox-system
----
-apiVersion: v1
-kind: Namespace
-metadata:
-  name: opensandbox

platform/sandbox-opensandbox/repository.yaml

@@ -1,3 +1,4 @@
+---
 apiVersion: source.toolkit.fluxcd.io/v1
 kind: GitRepository
 metadata:
@@ -6,6 +7,5 @@ metadata:
 spec:
   interval: 1h
   ref:
-    tag: helm/opensandbox/0.2.2
+    commit: 8f01e935c2cabba778cf37a152033fae062fa0f4
-  timeout: 60s
+  url: https://github.com/opensandbox-group/OpenSandbox.git
-  url: https://github.com/alibaba/OpenSandbox.git

platform/sandbox-opensandbox/template.yaml

@@ -0,0 +1,17 @@
+---
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: opensandbox-batchsandbox-template
+  namespace: opensandbox-system
+data:
+  batchsandbox-template.yaml: |
+    metadata:
+      labels:
+        ddupan.top/workload-class: sandbox
+    spec:
+      replicas: 1
+      template:
+        spec:
+          restartPolicy: Never
+          terminationGracePeriodSeconds: 30

platform/sandbox-opensandbox/values.yaml

@@ -1,41 +1,52 @@
+---
 apiVersion: v1
 kind: ConfigMap
 metadata:
   name: opensandbox-values
   namespace: opensandbox-system
+  labels:
+    reconcile.fluxcd.io/watch: Enabled
 data:
   values.yaml: |
     opensandbox-controller:
       controller:
-        logLevel: info
-        replicaCount: 1
         metrics:
           enabled: true
-          secure: false
           port: 8080
-        resources:
+          secure: false
-          requests:
-            cpu: 25m
-            memory: 64Mi
-          limits:
-            cpu: 500m
-            memory: 256Mi
 
     opensandbox-server:
       server:
         replicaCount: 1
+        env:
+          - name: OPENSANDBOX_SERVER_API_KEY
+            valueFrom:
+              secretKeyRef:
+                name: opensandbox-api-key
+                key: api-key
         resources:
-          requests:
-            cpu: 100m
-            memory: 256Mi
           limits:
             cpu: "1"
             memory: 1Gi
+          requests:
+            cpu: 100m
+            memory: 256Mi
+        volumeMounts:
+          - name: batchsandbox-template
+            mountPath: /etc/opensandbox/batchsandbox-template.yaml
+            subPath: batchsandbox-template.yaml
+            readOnly: true
+        volumes:
+          - name: batchsandbox-template
+            configMap:
+              name: opensandbox-batchsandbox-template
+
       configToml: |
         [server]
         host = "0.0.0.0"
         port = 80
         api_key = ""
+        max_sandbox_timeout_seconds = 86400
 
         [log]
         level = "INFO"
@@ -44,6 +55,10 @@ data:
         type = "kubernetes"
         execd_image = "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/execd:v1.0.22"
 
+        [storage]
+        allowed_host_paths = []
+        volume_default_size = "1Gi"
+
         [kubernetes]
         kubeconfig_path = ""
         namespace = "opensandbox"
@@ -52,11 +67,17 @@ data:
         informer_watch_timeout_seconds = 60
         snapshot_create_timeout_seconds = 900
         workload_provider = "batchsandbox"
-        batchsandbox_template_file = "/etc/opensandbox/example.batchsandbox-template.yaml"
+        image_pull_policy = "IfNotPresent"
+        batchsandbox_template_file = "/etc/opensandbox/batchsandbox-template.yaml"
 
         [egress]
         image = "sandbox-registry.cn-zhangjiakou.cr.aliyuncs.com/opensandbox/egress:v1.1.6"
         mode = "dns+nft"
+        disable_ipv6 = true
+
+        [secure_runtime]
+        type = "kata"
+        k8s_runtime_class = "kata-clh-runtime-rs"
 
     opensandbox-node-agent:
       enabled: false

platform/spire/README.md

@@ -74,6 +74,12 @@ Agent 已启用 Unix workload attestor，并为本机用户 `panxiao81`（UID `1
 export SPIFFE_ENDPOINT_SOCKET=unix:///run/spire/agent-sockets/spire-agent.sock
 ```
 
+宿主机已安装与 Agent Pod 同版本的 `/usr/local/bin/spire-agent` 1.15.3，供本地进程从
+Workload API 获取 JWT-SVID。二进制来自 SPIRE 官方 `linux-amd64-musl` release，安装时
+核对 tarball SHA-256
+`ca1a4d1155317bdd2afc7f36663828a10410c7c840e54725b90b4064b0a301c7`。升级 chart 时应
+同步升级这个 CLI 并重新核对官方 checksum，不能长期混用版本。
+
 该身份仅按 Unix UID 匹配，不是 SPIRE admin，也不会匹配 `sudo` 后以 root 运行的
 进程。`ClusterStaticEntry.spec.parentID` 绑定当前 `laptop` Kubernetes node UID；若
 节点被删除后重建，需从 `spire-server agent list` 取得新 Agent ID 并同步更新该字段。

platform/spire/RUNBOOK.md

@@ -274,14 +274,24 @@ CI job/Agent 不应接收长期 `BAO_TOKEN`。标准启动顺序是：
 5. 在同一进程树中以环境变量调用 `tofu`、Ansible 或其他工具；
 6. cleanup 尝试 `revoke-self`，随后销毁 job/VM/容器。
 
-通用 credential-exec 包装器未来应负责步骤 3–6。它必须满足：
+Gitea Actions 可使用
+[`panxiao81/ci-actions/spiffe-openbao-login@v1`](https://git.ddupan.top/panxiao81/ci-actions/src/tag/v1/spiffe-openbao-login)
+完成步骤 3、4 和 cleanup 时的 `revoke-self`。runner 只负责提供 Node.js 20、
+`spire-agent` 与 Workload API socket；role、audience 及是否登录由 workflow 明确声明，
+目标侧 Bao policy 仍负责最终授权。该 Action 必须满足：
 
 - 不把 JWT-SVID 或 Bao token 写到 stdout/stderr；
 - 不把凭据传入命令行参数，避免出现在进程列表；
-- 子进程退出后清除环境和临时文件；
+- job 结束后销毁 runner 及其 Actions 临时文件；
 - 不尝试把短期 token 上传到 Actions Secret 或 artifact；
 - role、audience 和目标命令由受审查的 pipeline 配置决定。
 
+Action 会按 GitHub Actions 协议把短期 token 写入 `GITHUB_ENV` 和 `GITHUB_STATE`，因此
+只允许用于一次性 Pod/VM runner，不能用于共享或持久 runner。只读取 Nexus public
+repository 时不需要 Bao 登录，可直接使用
+[`panxiao81/ci-actions/setup-nexus@v1`](https://git.ddupan.top/panxiao81/ci-actions/src/tag/v1/setup-nexus)
+配置 Ansible Galaxy、Go module proxy 与 OCI endpoint。
+
 Kubernetes 以外的执行环境不能伪造 ServiceAccount。未来应分别使用 host SPIRE
 Agent、TPM/DevID、cloud instance identity、GitHub OIDC 等初始证明接入同一信任模型。
 

platform/spire/values.yaml

@@ -43,6 +43,10 @@ spire-server:
       externalSecret:
         name: spire-external-kubeconfigs
         key: sandbox-controller
+    sandbox-runner-bundle:
+      externalSecret:
+        name: spire-external-kubeconfigs
+        key: sandbox-controller
   nodeAttestor:
     externalK8sPSAT:
       enabled: true
@@ -55,12 +59,19 @@ spire-server:
           kubeConfigName: sandbox
           serviceAccountAllowList:
             - spire-smoke:spire-smoke
+            - opensandbox:gitea-ci
           usePodUIDForAgentID: true
   externalControllerManagers:
     enabled: true
     clusters:
       sandbox:
         kubeConfigName: sandbox-controller
+        # Dynamic OpenSandbox runner identities are exact Pod-UID-bound
+        # ClusterStaticEntries created in the sandbox cluster. The chart
+        # defaults this reconciler to false, so enable the central registration
+        # path explicitly.
+        reconcile:
+          clusterStaticEntries: true
   bundlePublisher:
     externalK8sConfigMap:
       enabled: true
@@ -71,6 +82,12 @@ spire-server:
           configMapName: spire-bundle
           configMapKey: bundle.spiffe
           format: spiffe
+        sandbox-runner-bundle:
+          kubeConfigName: sandbox-runner-bundle
+          namespace: opensandbox
+          configMapName: spire-bundle-pem
+          configMapKey: bundle.pem
+          format: pem
   persistence:
     # PostgreSQL stores registrations, but the disk KeyManager still needs durable
     # storage for the trust-domain signing keys.