Azure DevOps自建Linux代理无法运行Terraform问题求助

问题现象 #

TerraformInstaller@1任务成功安装Terraform（版本1.7.5），但后续TerraformTaskV4@4执行init命令时，报错提示Error: Failed to find terraform tool in paths。该流水线在微软托管代理上可正常运行。

流水线代码 #

trigger:
  none
pr:
  '*'
pool:
  name: Self-Hosted-Agent(pool name)
  demands:
    agent.name -equals xxx(agent name)
stages:
  - stage: terraform_installation
    jobs:
      - job: installation
        steps:
          - task: TerraformInstaller@1
            displayName: 'Terraform installation'
            inputs:
              terraformVersion: 'latest'
  - stage: terraform_init
    jobs:
      - job: init
        steps:
          - checkout: self
          - task: TerraformTaskV4@4
            displayName: 'Terraform init'
            inputs:
              provider: 'azurerm'
              command: 'init'
              workingDirectory: ''
              environmentServiceNameAzureRM: ''
              backendServiceArm: ''
              backendAzureRmResourceGroupName: ''
              backendAzureRmStorageAccountName: ''
              backendAzureRmContainerName: ''
              backendAzureRmKey: '**'

Terraform安装成功日志 #

Starting: Terraform installation
Task : Terraform tool installer
Description : Find in cache or download a specific version of Terraform and prepend it to the PATH
Version : 1.216.24
Author : Microsoft Corporation
Help : Learn more about this task

Getting latest Terraform version.
Downloading: https://releases.hashicorp.com/terraform/1.7.5/terraform_1.7.5_linux_amd64.zip
Extracting archive
/usr/bin/unzip /package/_work/_temp/terraform-1.7.5-90331641-692a-439d-9d2d-eadd8ae50420.zip
Archive: /package/_work/_temp/terraform-1.7.5-90331641-692a-439d-9d2d-eadd8ae50420.zip
inflating: terraform
Caching tool: terraform 1.7.5 x64
Prepending PATH environment variable with directory: /package/_work/_tool/terraform/1.7.5/x64
Verifying Terraform installation…
/package/_work/_tool/terraform/1.7.5/x64/terraform version
Terraform v1.7.5
on linux_amd64
Finishing: Terraform installation

Terraform Init失败日志 #

Starting: Terraform init
Task : Terraform
Description : Execute terraform commands to manage resources on AzureRM, Amazon Web Services(AWS) and Google Cloud Platform(GCP)
Version : 4.227.24
Author : Microsoft Corporation
Help : Learn more about this task

##[error]Error: Failed to find terraform tool in paths
Finishing: Terraform init

已尝试操作 #

• 手动在Linux代理机器上将/package/_work/_tool/terraform/1.7.5/x64加入系统PATH环境变量
• 放行所有HashiCorp和Terraform相关URL

解决方法 #

方法1：合并安装与Init步骤到同一Job #

Azure DevOps中不同Job的环境变量相互隔离，第一个Job中TerraformInstaller设置的PATH无法传递到第二个Job。将两个步骤合并到同一Job即可解决：

trigger:
  none
pr:
  '*'
pool:
  name: Self-Hosted-Agent(pool name)
  demands:
    agent.name -equals xxx(agent name)
stages:
  - stage: terraform_workflow
    jobs:
      - job: terraform_job
        steps:
          - checkout: self
          - task: TerraformInstaller@1
            displayName: 'Terraform installation'
            inputs:
              terraformVersion: 'latest'
          - task: TerraformTaskV4@4
            displayName: 'Terraform init'
            inputs:
              provider: 'azurerm'
              command: 'init'
              workingDirectory: ''
              environmentServiceNameAzureRM: ''
              backendServiceArm: ''
              backendAzureRmResourceGroupName: ''
              backendAzureRmStorageAccountName: ''
              backendAzureRmContainerName: ''
              backendAzureRmKey: '**'

方法2：在Init Job中手动注入PATH #

若必须分Job执行，可在Init步骤前添加脚本，将Terraform路径注入当前Job的环境变量：

- stage: terraform_init
  jobs:
    - job: init
      steps:
        - checkout: self
        - script: |
            echo "##vso[task.setvariable variable=PATH]/package/_work/_tool/terraform/1.7.5/x64:$PATH"
          displayName: 'Add Terraform to PATH'
        - task: TerraformTaskV4@4
          displayName: 'Terraform init'
          inputs:
            provider: 'azurerm'
            command: 'init'
            # 其余参数保持不变

方法3：在代理机器全局安装Terraform #

跳过TerraformInstaller任务，直接在Linux代理机器上全局安装指定版本的Terraform，并确保其路径加入系统默认PATH，所有Job即可直接调用。

内容的提问来源于stack exchange，提问作者SAUPARNA DATTA