前言
相信经过前一篇文章的学习，大家已经对Helm有所了解，本篇文章介绍另一款工具 Kustomize，为什么Helm如此流行，还会出现 Kustomize？而且 Kustomize 自 kubectl 1.14以来早已内置到其中，说明官方对其很认可。
我认为还是Helm 参数化模板方式来自定义配置需要学习复杂的DSL语法，难上手，易出错，而 kustomize 是使用 Kubernetes 原生概念帮助用户创作并复用声明式配置。
认识 Kustomize
安装
brew install kustomize示例解析
demo-manifests ├── base │ ├── deployment.yaml │ ├── kustomization.yaml │ ├── service-account.yaml │ └── service.yaml └── services ├── demo-app │ ├── _common │ │ ├── deployment-patch.yaml │ │ ├── kustomization.yaml │ │ └── namespace.yaml │ ├── dev │ │ ├── config.json │ │ ├── deployment-patch.yaml │ │ └── kustomization.yaml │ ├── staging │ │ ├── config.json │ │ ├── deployment-patch.yaml │ │ └── kustomization.yaml │ └── prod │ │ ├── config.json │ │ ├── deployment-patch.yaml │ │ └── kustomization.yaml └── demo-app2 └── xxx
先看base 目录，有几个常见的deployment/service/service-account YAML 资源文件
还有个 kustomization.yaml 配置文件
这里面包含了刚才的几个resource文件以及要应用于它们的一些自定义，如添加一个通用的标签Common Labels
这时候也可以通过 kustomize build 命令来生成完整的YAML进行查看
kustomize build demo-manifests/base > base.yaml
base.yaml
build 出来的 YAML 每个资源对象上都会存在通用的标签 managed-by: Kustomize
接下来看 Service目录，此目录存放所有的服务项目，比如demo-app ， 里面的YAML就是来覆盖base 也就是官方说的 Overlays
只需要把不同的资源描述通过Patch方式覆盖掉base中的就行了。这边我又将三个环境 dev/staging/prod 公共的部分抽取出来放入 common文件夹。
_common/kustomization.yaml文件如下
apiVersion: kustomize.config.k8s.io/v1beta1 kind: Kustomization resources: - ../../../base - namespace.yaml patchesStrategicMerge: - deployment-patch.yaml
我这个demo-app 需要configMap，另外心跳接口也不一样， 所以 deployment-patch.yaml 需要写出这些不一样的，然后去覆盖base中的。
apiVersion: apps/v1 kind: Deployment metadata: name: NAME_PLACEHOLDER spec: template: spec: serviceAccountName: NAME_PLACEHOLDER containers: - name: app image: wadexu007/demo:IMG_TAG_PLACEHOLDER livenessProbe: failureThreshold: 5 httpGet: path: /pizzas port: 8080 initialDelaySeconds: 10 periodSeconds: 40 timeoutSeconds: 1 readinessProbe: failureThreshold: 5 httpGet: path: /pizzas port: 8080 initialDelaySeconds: 10 periodSeconds: 20 timeoutSeconds: 1 volumeMounts: - name: config-volume mountPath: /app/conf/config.json subPath: config.json volumes: - name: config-volume configMap: name: demo-app-config
deployment-patch.yaml
--- apiVersion: v1 kind: Namespace metadata: name: demo
namespace.yaml
### 文章首发于博客园 https://www.cnblogs.com/wade-xu/p/16839829.html
最后我们看 dev/kustomization.yaml
apiVersion: kustomize.config.k8s.io/v1beta1 kind: Kustomization resources: - ../_common
namespace: demo
commonLabels: app: demo-app
replicas: - count: 1 name: demo-app configMapGenerator: - files: - config.json name: demo-app-config patches: - patch: |- - op: replace path: /metadata/name value: demo-app target: name: NAME_PLACEHOLDER
patchesStrategicMerge: - deployment-patch.yaml
images: - name: wadexu007/demo newTag: 1.0.0
dev env 里面将replicas设置成1， 用config.json 生成 configMap
{ "SOME_CONFIG": "/demo/path" }
config.json
deployment-patch.yaml 里面也将container resource request/limit 配置设置小一点， 采用 patchesStrategicMerge 与 common 和 base里面的deployment资源合并。
apiVersion: apps/v1 kind: Deployment metadata: name: NAME_PLACEHOLDER spec: template: spec: containers: - name: app resources: limits: cpu: 1 memory: 1Gi requests: cpu: 200m memory: 256Mi
deployment-patch.yaml
另外还采用了 patch 的方式，改了每个资源的name --> demo-app
以及images 替换了原来的 image name and tag.
kustomize 命令如下
kustomize build demo-manifests/services/demo-app/dev > demo-app.yaml
如果用 kubectl 命令的话，示例如下
kubectl kustomize services/demo-app/dev/ > demo-app.yaml
最终结果
apiVersion: v1 kind: Namespace metadata: labels: app: demo-app name: demo --- apiVersion: v1 kind: ServiceAccount metadata: labels: app: demo-app managed-by: Kustomize name: demo-app namespace: demo --- apiVersion: v1 data: config.json: |- { "SOME_CONFIG": "/demo/path" } kind: ConfigMap metadata: labels: app: demo-app name: demo-app-config-t7c64mbtt2 namespace: demo --- apiVersion: v1 kind: Service metadata: labels: app: demo-app managed-by: Kustomize name: demo-app namespace: demo spec: ports: - port: 8080 protocol: TCP targetPort: http selector: app: demo-app managed-by: Kustomize type: ClusterIP --- apiVersion: apps/v1 kind: Deployment metadata: labels: app: demo-app managed-by: Kustomize name: demo-app namespace: demo spec: replicas: 1 selector: matchLabels: app: demo-app managed-by: Kustomize template: metadata: labels: app: demo-app managed-by: Kustomize spec: containers: - image: wadexu007/demo:1.0.0 imagePullPolicy: IfNotPresent livenessProbe: failureThreshold: 5 httpGet: path: /pizzas port: 8080 initialDelaySeconds: 10 periodSeconds: 40 timeoutSeconds: 1 name: app ports: - containerPort: 8080 name: http readinessProbe: failureThreshold: 5 httpGet: path: /pizzas port: 8080 initialDelaySeconds: 10 periodSeconds: 20 timeoutSeconds: 1 resources: limits: cpu: 1 memory: 1Gi requests: cpu: 200m memory: 256Mi securityContext: allowPrivilegeEscalation: false volumeMounts: - mountPath: /app/conf/config.json name: config-volume subPath: config.json serviceAccountName: demo-app volumes: - configMap: name: demo-app-config-t7c64mbtt2 name: config-volume
demo-app.yaml
### https://www.cnblogs.com/wade-xu/p/16839829.html
staging/prod 文件夹也是类似dev的文件，环境不一样，config.json 以及资源配置，image tag 不一样，显示申明就行了。
完整代码请参考我的 Github
涵盖 Kustomize 知识点
- commonLabels
- patchesStrategicMerge
- patches
- configMapGenerator
- replicas
- images
总结
经过本篇文章的学习，以及上一篇 [云原生之旅 - 5）Kubernetes时代的包管理工具 Helm ]关于Helm的学习，您已经能体会到两款工具的不同之处。
总的来说，Helm 有自己一套体系来管理应用，而 kustomize 更轻量级，也更灵活。另外，Kustomize也有 Terraform provider 通过TF来安装。
