Merge pull request 51: 部署 SPIRE workload identity 基础设施
yaml / yaml (push) Successful in 18s

This commit was merged in pull request #51.
This commit is contained in:
2026-09-13 15:35:53 +00:00
10 changed files with 302 additions and 0 deletions
+2
View File
@@ -50,4 +50,6 @@ sudo k3s kubectl -n flux-system get gitrepositories,kustomizations
- VictoriaMetrics Operator 已固定现有 chart `0.66.2` 并完成分阶段 Flux HelmRelease
接管;Metrics、Logs、Traces 与 Grafana 也已统一完成 Flux 接管;
- External Secrets Operator 已固定 chart `2.8.0` 并完成分阶段接管;
- SPIRE 已按官方 hardened chart `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;
- root Kustomization 与所有 brownfield 子 Kustomization 继续保持 `prune: false`。
+14
View File
@@ -0,0 +1,14 @@
apiVersion: kustomize.toolkit.fluxcd.io/v1
kind: Kustomization
metadata:
name: spire
namespace: flux-system
spec:
interval: 10m
path: ./platform/spire
prune: false
sourceRef:
kind: GitRepository
name: flux-system
timeout: 15m
wait: false
+1
View File
@@ -10,4 +10,5 @@ resources:
- apps/gitea-actions.yaml
- apps/http-echo.yaml
- apps/openebs.yaml
- apps/spire.yaml
- apps/observability.yaml
+80
View File
@@ -0,0 +1,80 @@
# SPIFFE/SPIRE workload identity
SPIRE 是 homelab 的机器与 workload identity 根。人类身份继续由 Samba AD 与
Authelia 提供;SPIRE 不替代人类 OIDC,也不承担目标服务的资源授权。
## 部署范围
Flux 安装 SPIFFE hardened charts:
- `spire-crds` `0.6.1`;
- `spire` `0.30.2`(SPIRE `1.15.3`);
- SPIRE Server、Agent、Controller Manager、SPIFFE CSI Driver;
- OIDC Discovery Provider。
未启用 Tornjak、SPIRE Identity Exchange、SPIKE、federation、Delegated Identity
API 或 Broker API。Trust domain 是 `ddupan.top`,Kubernetes cluster name 是
`homelab`。
## PostgreSQL bootstrap
SPIRE registration datastore 使用共享 CloudNativePG:
```text
host: shared-postgresql-rw.shared-db.svc.cluster.local:5432
database: spire
role: spire
```
数据库与 role 当前是手工创建的临时 bootstrap。密码只存在于
`spire-server/spire-postgresql` Secret 的 `password` key 中,不提交到 Git。
在 PostgreSQL tenant operator/DBaaS 接管前,不得删除该 Secret 或重置数据库
role 密码。
后续声明式管理必须保持这一 Secret 接口,或者在同一个变更中更新
`spire-server.dataStore.sql.externalSecret`,避免数据库凭据出现两个写入方。
PostgreSQL保存 registration state;SPIRE Server 的 disk KeyManager 仍使用一个
`1Gi`、`localpv-zfs-ceph` PVC 保存 trust-domain signing keys。数据库备份不能替代
该 PVC/密钥的备份。
## 身份签发策略
默认的全 Pod fallback `ClusterSPIFFEID` 已关闭。新增 workload 必须显式创建
`ClusterSPIFFEID`,并以 namespace、ServiceAccount、Pod label 等 selector 收窄。
不得仅因 Pod 能挂载 CSI socket 就给它签发身份。
稳定的 JWT issuer 预留为:
```text
https://spire-oidc.ad.ddupan.top
```
首次部署时 OIDC Discovery Provider 仅为 ClusterIP,内部为明文 HTTP;尚未创建
外部 DNS 或 Gateway route。接入 OpenBao 前,在现有 Envoy Gateway 上终止 TLS、
增加内网 DNS,并验证 discovery document 的 `issuer` 与上述 URL 完全一致。
## 首次部署与验证
合并后观察:
```bash
sudo k3s kubectl -n flux-system get kustomization spire
sudo k3s kubectl -n spire-mgmt get helmrelease
sudo k3s kubectl -n spire-server get pods,pvc
sudo k3s kubectl -n spire-system get daemonset,pods
```
必须先确认 `spire-crds` Ready,随后 `spire` Ready。SPIRE Server 应连接 PostgreSQL,
Agent 应通过 PSAT attestation 注册,CSI Driver 应在节点 Ready。
首个业务验收另行增加一个专用测试 Pod 与 `ClusterSPIFFEID`,验证取得
`aud=openbao` 的 JWT-SVID 后登录 OpenBao。PoC 完成前不修改生产认证方式。
## 恢复边界
- 恢复顺序:共享 PostgreSQL、SPIRE Server signing-key PVC、SPIRE Server、Agent;
- issuer URL 与 trust domain 初始化后不得随意修改;
- 丢失 signing keys 会使既有 SVID 和下游 JWKS 信任失效;
- PostgreSQL或 SPIRE 不可用时,不得用新的空数据库覆盖现有状态;
- 当前 Flux root 与本 Kustomization 均保持 `prune: false`,删除资源需单独审计。
+30
View File
@@ -0,0 +1,30 @@
apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
name: spire-crds
namespace: spire-mgmt
spec:
chart:
spec:
chart: spire-crds
interval: 1h
sourceRef:
kind: HelmRepository
name: spiffe-hardened
version: 0.6.1
driftDetection:
mode: enabled
install:
crds: CreateReplace
strategy:
name: RetryOnFailure
retryInterval: 5m
interval: 30m
releaseName: spire-crds
targetNamespace: spire-mgmt
timeout: 10m
upgrade:
crds: CreateReplace
strategy:
name: RetryOnFailure
retryInterval: 5m
+34
View File
@@ -0,0 +1,34 @@
apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
name: spire
namespace: spire-mgmt
spec:
chart:
spec:
chart: spire
interval: 1h
sourceRef:
kind: HelmRepository
name: spiffe-hardened
version: 0.30.2
dependsOn:
- name: spire-crds
namespace: spire-mgmt
driftDetection:
mode: enabled
install:
strategy:
name: RetryOnFailure
retryInterval: 5m
interval: 30m
releaseName: spire
targetNamespace: spire-mgmt
timeout: 15m
upgrade:
strategy:
name: RetryOnFailure
retryInterval: 5m
valuesFrom:
- kind: ConfigMap
name: spire-values
+8
View File
@@ -0,0 +1,8 @@
apiVersion: source.toolkit.fluxcd.io/v1
kind: HelmRepository
metadata:
name: spiffe-hardened
namespace: spire-mgmt
spec:
interval: 1h
url: https://spiffe.github.io/helm-charts-hardened/
+16
View File
@@ -0,0 +1,16 @@
apiVersion: kustomize.config.k8s.io/v1beta1
kind: Kustomization
generatorOptions:
disableNameSuffixHash: true
labels:
reconcile.fluxcd.io/watch: Enabled
configMapGenerator:
- name: spire-values
namespace: spire-mgmt
files:
- values.yaml=values.yaml
resources:
- namespaces.yaml
- helmrepository.yaml
- helmrelease-crds.yaml
- helmrelease.yaml
+14
View File
@@ -0,0 +1,14 @@
apiVersion: v1
kind: Namespace
metadata:
name: spire-mgmt
---
apiVersion: v1
kind: Namespace
metadata:
name: spire-server
---
apiVersion: v1
kind: Namespace
metadata:
name: spire-system
+103
View File
@@ -0,0 +1,103 @@
global:
k8s:
clusterDomain: cluster.local
spire:
clusterName: homelab
trustDomain: ddupan.top
# This is a stable external identity even before its Gateway route is added.
# Changing the issuer after consumers trust it would invalidate their config.
jwtIssuer: https://spire-oidc.ad.ddupan.top
caSubject:
country: CN
organization: ddupan homelab
commonName: ddupan.top
namespaces:
create: false
system:
name: spire-system
server:
name: spire-server
recommendations:
enabled: true
namespaceLayout: true
namespacePSS: true
priorityClassName: true
strictMode: true
securityContexts: true
prometheus: false
spire-server:
kind: statefulset
replicaCount: 1
auditLogEnabled: true
persistence:
# PostgreSQL stores registrations, but the disk KeyManager still needs durable
# storage for the trust-domain signing keys.
type: pvc
size: 1Gi
accessMode: ReadWriteOnce
storageClass: localpv-zfs-ceph
dataStore:
sql:
databaseType: postgres
databaseName: spire
host: shared-postgresql-rw.shared-db.svc.cluster.local
port: 5432
username: spire
options:
- sslmode: require
externalSecret:
enabled: true
name: spire-postgresql
key: password
maxOpenConns: 20
maxIdleConns: 2
connMaxLifetime: 30m
controllerManager:
enabled: true
# Identities are opt-in. Do not issue a fallback SVID to every pod in the
# cluster merely because it can mount the Workload API socket.
identities:
clusterSPIFFEIDs:
default:
enabled: false
test-keys:
enabled: false
spire-agent:
nodeAttestor:
k8sPSAT:
enabled: true
workloadAttestors:
k8s:
enabled: true
unix:
enabled: false
spiffe-csi-driver:
enabled: true
spiffe-oidc-discovery-provider:
enabled: true
# TLS will terminate at the existing Envoy Gateway when the stable issuer is
# exposed. The service remains ClusterIP-only during the first deployment.
tls:
spire:
enabled: false
externalSecret:
enabled: false
certManager:
enabled: false
upstream:
enabled: false
tornjak-frontend:
enabled: false
spire-identity-exchange:
enabled: false
spike-keeper:
enabled: false
spike-nexus:
enabled: false
spike-pilot:
enabled: false