将Cloud Run镜像部署到Kubernetes集群：变量替换失败求助

问题描述 #

已通过cloudbuild.yaml成功构建Docker镜像，但使用deployment.yaml部署到GKE集群时，镜像地址中的$PROJECT_ID、${_DEPLOYMENT_TYPE}、$COMMIT_SHA变量未被正确替换，触发以下报错：

Error: failed to prepare deployment: failed to update container of objects: failed to parse reference from image "us-west1-docker.pkg.dev/$PROJECT_ID/abc/abc-server-${_DEPLOYMENT_TYPE}:$COMMIT_SHA": could not parse reference: us-west1-docker.pkg.dev/$PROJECT_ID/abc/abc-server-${_DEPLOYMENT_TYPE}:$COMMIT_SHA

根本原因 #

Cloud Build仅会替换cloudbuild.yaml自身参数中的变量，不会自动修改deployment.yaml文件内的占位符；而gke-deploy的--image参数作用是直接替换部署配置中的镜像地址，并非解析文件内的变量。原配置中deployment.yaml的镜像地址始终保留未替换的变量，导致Kubernetes无法解析无效的镜像引用。

解决方案 #

方案1：利用gke-deploy自动替换镜像地址 #

修改deployment.yaml中的镜像地址为任意占位值，让gke-deploy通过--image参数自动替换该字段：

spec:
  containers:
    - name: idr-server-k8ts-test
      image: placeholder-image # 占位值，会被gke-deploy自动替换
      ports:
        - containerPort: 8080

cloudbuild.yaml中的部署步骤保持不变，--image参数已正确拼接变量后的镜像地址，gke-deploy会自动找到deployment中的容器并替换镜像地址。

方案2：先替换deployment.yaml变量再部署 #

在Cloud Build步骤中添加变量替换操作，先处理deployment.yaml中的占位符，再执行部署：

1. 添加变量替换步骤：

- id: "替换deployment变量"
  name: "alpine"
  args: ["sh", "-c", "envsubst < deployment.yaml > deployment-temp.yaml"]
  env:
    - "PROJECT_ID=$PROJECT_ID"
    - "_DEPLOYMENT_TYPE=${_DEPLOYMENT_TYPE}"
    - "COMMIT_SHA=$COMMIT_SHA"

2. 修改部署步骤的配置文件路径为替换后的临时文件：

- id: "deploy container image to GKE"
  name: "gcr.io/cloud-builders/gke-deploy"
  args: [
    "--cluster", "${_CLUSTER_NAME}",
    "run",
    "--filename", "deployment-temp.yaml",
    "--location", "${_COMPUTE_ZONE}"
  ]

内容的提问来源于stack exchange，提问作者Salad.Guyo