使用 Open Policy Agent 实现可信镜像仓库检查

从互联网（或可信镜像仓库库以外的任何地方）拉取未知镜像会带来风险——例如恶意软件。但是还有其他很好的理由来维护单一的可信来源，例如在企业中实现可支持性。通过确保镜像仅来自受信任的镜像仓库，可以密切控制镜像库存，降低软件熵和蔓延的风险，并提高集群的整体安全性。除此以外，有时还会需要检查镜像的 tag，比如禁止使用 latest 镜像。

这今天我们尝试用“策略即代码”的实现 OPA 来实现功能。

还没开始之前可能有人会问：明明可以实现个 Admission Webhook^[1] 就行，为什么还要加上 OPA？

确实可以，但是这样策略和逻辑都会耦合在一起，当策略需要调整的时候需要修改代码重新发布。而 OPA 就是用来做解耦的，其更像是一个策略的执行引擎。

什么是 OPA

Open Policy Agent（以下简称 OPA，发音 “oh-pa”）一个开源的通用策略引擎，可以统一整个堆栈的策略执行。OPA 提供了一种高级声明性语言（Rego），可让你将策略指定为代码和简单的 API，以从你的软件中卸载策略决策。你可以使用 OPA 在微服务、Kubernetes、CI/CD 管道、API 网关等中实施策略。

Rego 是一种高级的声明性语言，是专门为 OPA 建立的。更多 OPA 的介绍可以看 Open Policy Agent 官网^[2]，不想看英文直接看这里^[3]。

2021-07-09-18-26-25

现在进入正题。

启动集群

启动 minikube

minikube start

创建用于部署 OPA 的命名空间

创建并切换到命名空间 opa （命名空间的切换使用 kubens，更多工具介绍见这里）

kubectl create namespace opakubens opa

在 Kubernetes 上部署 OPA

Kubernetes 和 OPA 间的通信必须使用 TLS 进行保护。配置 TLS，使用 openssl 创建证书颁发机构（certificate authority CA）和 OPA 的证书/秘钥对。

openssl genrsa -out ca.key 2048openssl req -x509 -new -nodes -key ca.key -days 100000 -out ca.crt -subj "/CN=admission_ca"

为 OPA 创建 TLS 秘钥和证书：

cat >server.conf <<EOF[req]req_extensions = v3_reqdistinguished_name = req_distinguished_nameprompt = no[req_distinguished_name]CN = opa.opa.svc[ v3_req ]basicConstraints = CA:FALSEkeyUsage = nonRepudiation, digitalSignature, keyEnciphermentextendedKeyUsage = clientAuth, serverAuthsubjectAltName = @alt_names[alt_names]DNS.1 = opa.opa.svcEOF

注意 CN 和 alt_names 必须与后面创建 OPA service 的匹配。

openssl genrsa -out server.key 2048openssl req -new -key server.key -out server.csr -config server.confopenssl x509 -req -in server.csr -CA ca.crt -CAkey ca.key -CAcreateserial -out server.crt -days 100000 -extensions v3_req -extfile server.conf

为 OPA 创建保存 TLS 凭证的 Secret：

kubectl create secret tls opa-server --cert=server.crt --key=server.key

将 OPA 部署为准入控制器（admission controller）。

admission-controller.yaml：

# 授权 OPA/kube-mgmt 对资源的只读权限# kube-mgmt 会同步资源信息给 OPA，以便在策略中使用kind: ClusterRoleBindingapiVersion: rbac.authorization.k8s.io/v1metadata: name: opa-viewerroleRef: kind: ClusterRole name: view apiGroup: rbac.authorization.k8s.iosubjects:- kind: Group name: system:serviceaccounts:opa apiGroup: rbac.authorization.k8s.io---# 为 OPA/kube-mgmt 定义角色来在 configmaps 中更新策略状态kind: RoleapiVersion: rbac.authorization.k8s.io/v1metadata: namespace: opa name: configmap-modifierrules:- apiGroups: [""] resources: ["configmaps"] verbs: ["update", "patch"]---# 为 OPA/kube-mgmt 授予角色kind: RoleBindingapiVersion: rbac.authorization.k8s.io/v1metadata: namespace: opa name: opa-configmap-modifierroleRef: kind: Role name: configmap-modifier apiGroup: rbac.authorization.k8s.iosubjects:- kind: Group name: system:serviceaccounts:opa apiGroup: rbac.authorization.k8s.io---kind: ServiceapiVersion: v1metadata: name: opa namespace: opaspec: selector: app: opa ports: - name: https protocol: TCP port: 443 targetPort: 8443---apiVersion: apps/v1kind: Deploymentmetadata: labels: app: opa namespace: opa name: opaspec: replicas: 1 selector: matchLabels: app: opa template: metadata: labels: app: opa name: opa spec: containers: # WARNING: OPA is NOT running with an authorization policy configured. This # means that clients can read and write policies in OPA. If you are # deploying OPA in an insecure environment, be sure to configure # authentication and authorization on the daemon. See the Security page for # details: https://www.openpolicyagent.org/docs/security.html. - name: opa image: openpolicyagent/opa:0.30.1-rootless args: - "run" - "--server" - "--tls-cert-file=/certs/tls.crt" - "--tls-private-key-file=/certs/tls.key" - "--addr=0.0.0.0:8443" - "--addr=http://127.0.0.1:8181" - "--log-format=json-pretty" - "--set=decision_logs.console=true" volumeMounts: - readOnly: true mountPath: /certs name: opa-server readinessProbe: httpGet: path: /health?plugins&bundle scheme: HTTPS port: 8443 initialDelaySeconds: 3 periodSeconds: 5 livenessProbe: httpGet: path: /health scheme: HTTPS port: 8443 initialDelaySeconds: 3 periodSeconds: 5 - name: kube-mgmt image: openpolicyagent/kube-mgmt:0.11 args: - "--replicate=v1/pods" volumes: - name: opa-server secret: secretName: opa-server---kind: ConfigMapapiVersion: v1metadata: name: opa-default-system-main namespace: opadata: main: | package system import data.kubernetes.validating.images main = { "apiVersion": "admission.k8s.io/v1beta1", "kind": "AdmissionReview", "response": response, } default uid = "" uid = input.request.uid response = { "allowed": false, "uid": uid, "status": { "reason": reason, }, } { reason = concat(", ", images.deny) reason != "" } else = {"allowed": true, "uid": uid}

kubectl apply -f admission-controller.yaml

接下来，生成将用于将 OPA 注册为准入控制器的 manifest。

cat > webhook-configuration.yaml <<EOFkind: ValidatingWebhookConfigurationapiVersion: admissionregistration.k8s.io/v1beta1metadata: name: opa-validating-webhookwebhooks: - name: validating-webhook.openpolicyagent.org rules: - operations: ["CREATE", "UPDATE"] apiGroups: ["*"] apiVersions: ["*"] resources: ["pods"] clientConfig: caBundle: $(cat ca.crt | base64 | tr -d '\n') service: namespace: opa name: opaEOF

生成的配置文件包含 CA 证书的 base64 编码，以便可以在 Kubernetes API 服务器和 OPA 之间建立 TLS 连接。

kubectl apply -f webhook-configuration.yaml

查看 OPA 日志：

kubectl logs -l app=opa -c opa -f

定义策略并通过 Kubernetes 将其加载到 OPA

这里我们定义了对容器镜像的检查：

•是否来自受信任的仓库•是否使用了 latest tag 的镜像

image-policy.rego

package kubernetes.validating.imagesdeny[msg] { some i input.request.kind.kind == "Pod" image := input.request.object.spec.containers[i].image endswith(image, ":latest") msg := sprintf("Image '%v' used latest image", [image]) } { some i input.request.kind.kind == "Pod" image := input.request.object.spec.containers[i].image not startswith(image, "192.168.64.1:5000") msg := sprintf("Image '%v' comes from untrusted registry", [image])}

kubectl create configmap image-policy --from-file=image-policy.rego

检查 configmap 的 annotation openpolicyagent.org/policy-status 值是否 为 '{"status":"ok"}'。否则，就要根据报错信息处理问题。

注：192.168.64.1:5000 是笔者本地容器运行的一个私有仓库。

version: '3.6'services: registry: image: registry:2.7.1 container_name: registry restart: always environment: REGISTRY_HTTP_ADDR: 0.0.0.0:5000 REGISTRY_STORAGE: filesystem REGISTRY_STORAGE_DELETE_ENABLED: 'true' REGISTRY_STORAGE_FILESYSTEM_ROOTDIRECTORY: /var/lib/registry ports: - '5000:5000' volumes: - '/Users/addo/Downloads/tmp/registry:/var/lib/registry'

测试

pod-bad-repo.yaml

apiVersion: v1kind: Podmetadata: creationTimestamp: null labels: run: web-server name: web-serverspec: containers: - image: nginx:1.21.1 name: web-server resources: {} dnsPolicy: ClusterFirst restartPolicy: Alwaysstatus: {}

kubectl apply -f pod-bad-repo.yamlError from server (Image 'nginx:1.21.1' comes from untrusted registry): error when creating "pod-bad-repo.yaml": admission webhook "validating-webhook.openpolicyagent.org" denied the request: Image 'nginx:1.21.1' comes from untrusted registry

pod-bad-tag.yaml

apiVersion: v1kind: Podmetadata: creationTimestamp: null labels: run: web-server name: web-serverspec: containers: - image: 192.168.64.1:5000/nginx:latest name: web-server resources: {} dnsPolicy: ClusterFirst restartPolicy: Alwaysstatus: {}

kubectl apply -f pod-bad-tag.yamlError from server (Image '192.168.64.1:5000/nginx:latest' used latest image): error when creating "pod-bad-tag.yaml": admission webhook "validating-webhook.openpolicyagent.org" denied the request: Image '192.168.64.1:5000/nginx:latest' used latest image

pod-ok.yaml

apiVersion: v1kind: Podmetadata: creationTimestamp: null labels: run: web-server name: web-serverspec: containers: - image: 192.168.64.1:5000/nginx:1.21.1 name: web-server resources: {} dnsPolicy: ClusterFirst restartPolicy: Alwaysstatus: {}

kubectl apply -f pod-ok.yamlpod/web-server created

总结

策略即代码，以代码的实现表达策略；在通过策略与执行引擎的解耦分离，让策略更加的灵活。

后面我们再探索 OPA 的更多场景。

引用链接

[1] Admission Webhook: https://kubernetes.io/docs/reference/access-authn-authz/extensible-admission-controllers/
[2] Open Policy Agent 官网: https://www.openpolicyagent.org/
[3] 这里: https://cloudnative.to/blog/introducing-policy-as-code-the-open-policy-agent-opa/