GCP Cloud Build YAML能否重复使用步骤并传入不同文件名参数？

问题描述 #

想要在Cloud Build YAML中重复使用gcr.io/cloud-builders/gke-deploy步骤，仅通过修改--filename参数部署不同的K8s资源，但第二个步骤执行时报错：

Error: failed to prepare deployment: failed to save suggested configuration files to "output/suggested": output directory "output/suggested" exists and is not empty

原配置代码：

- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-job'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
  id: 'job'
- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-app'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
  waitFor: ['job']

报错原因 #

gke-deploy run命令默认会在当前工作目录下生成output/suggested目录，用于保存建议的K8s配置文件。第一个步骤执行后该目录已存在，第二个步骤尝试创建时因目录非空而报错。

解决方法 #

方法1：为每个步骤指定独立输出目录 #

通过--output-dir参数给每个gke-deploy步骤指定不同的输出目录，避免目录冲突：

- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-job'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
    - '--output-dir=output/job-suggested'
  id: 'job'
- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-app'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
    - '--output-dir=output/app-suggested'
  waitFor: ['job']

方法2：跳过生成建议配置文件 #

如果不需要保留建议的配置文件，直接添加--skip-suggest参数，命令将不再创建output/suggested目录：

- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-job'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
    - '--skip-suggest'
  id: 'job'
- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-app'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
    - '--skip-suggest'
  waitFor: ['job']

方法3：执行第二个步骤前清理目录 #

在第二个gke-deploy步骤前，添加一个清理目录的步骤，删除已存在的output/suggested：

- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-job'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
  id: 'job'
- name: 'gcr.io/cloud-builders/rm'
  args: ['-rf', 'output/suggested']
  waitFor: ['job']
- name: 'gcr.io/cloud-builders/gke-deploy'
  args:
    - run
    - '--filename=kubernetes-app'
    - '--image=asia-south1-docker.pkg.dev/product-staging-apps/test-stg-repo/staging-app:$SHORT_SHA'
    - '--location=asia-south1'
    - '--cluster=stg-gke-cluster'
  waitFor: ['rm']

内容的提问来源于stack exchange，提问作者Naveen