查看“GitLab 流水线架构”的源代码

构建流水线的主要方法有三种，每种方法都有自己的优势。如果需要，可以混合和匹配这些方法：

* 基本：适用于所有配置都集中在一个易于找到的地方的简单项目。
* 有向无环图：适用于需要高效执行的大型复杂项目。
* 父子流水线：适用于具有大量独立定义组件的 monorepos 和项目。

<br>

== 基本流水线 ==

这是极狐GitLab 中最简单的流水线。它同时运行构建阶段的所有内容，一旦所有这些都完成，它会以相同的方式运行测试阶段的所有内容，依此类推。 这不是最有效的，如果您有很多步骤，它会变得非常复杂，但更容易维护：

<br>

<pre>stages:
- build
- test
- deploy

image: alpine

build_a:
stage: build
script:
    - echo &quot;This job builds something.&quot;

build_b:
stage: build
script:
    - echo &quot;This job builds something else.&quot;

test_a:
stage: test
script:
    - echo &quot;This job tests something. It will only run when all jobs in the&quot;
    - echo &quot;build stage are complete.&quot;

test_b:
stage: test
script:
    - echo &quot;This job tests something else. It will only run when all jobs in the&quot;
    - echo &quot;build stage are complete too. It will start at about the same time as test_a.&quot;

deploy_a:
stage: deploy
script:
    - echo &quot;This job deploys something. It will only run when all jobs in the&quot;
    - echo &quot;test stage complete.&quot;
environment: production

deploy_b:
stage: deploy
script:
    - echo &quot;This job deploys something else. It will only run when all jobs in the&quot;
    - echo &quot;test stage complete. It will start at about the same time as deploy_a.&quot;
environment: production</pre>
<br>

== 有向无环图流水线 ==

如果效率对您很重要，并且您希望一切都尽可能快地运行，则可以使用有向无环图（DAG）。 使用 needs 关键字 定义作业之间的依赖关系。当系统知道您的作业之间的关系时，它可以尽可能快地运行所有内容，甚至在可能的情况下跳到后续阶段。

在下面的示例中，如果 build_a 和 test_a 比 build_b 和 test_b 快得多，即使 build_b 仍在运行，系统也会启动 deploy_a。

<br>

<pre>stages:
- build
- test
- deploy

image: alpine

build_a:
stage: build
script:
    - echo &quot;This job builds something quickly.&quot;

build_b:
stage: build
script:
    - echo &quot;This job builds something else slowly.&quot;

test_a:
stage: test
needs: [build_a]
script:
    - echo &quot;This test job will start as soon as build_a finishes.&quot;
    - echo &quot;It will not wait for build_b, or other jobs in the build stage, to finish.&quot;

test_b:
stage: test
needs: [build_b]
script:
    - echo &quot;This test job will start as soon as build_b finishes.&quot;
    - echo &quot;It will not wait for other jobs in the build stage to finish.&quot;

deploy_a:
stage: deploy
needs: [test_a]
script:
    - echo &quot;Since build_a and test_a run quickly, this deploy job can run much earlier.&quot;
    - echo &quot;It does not need to wait for build_b or test_b.&quot;
environment: production

deploy_b:
stage: deploy
needs: [test_b]
script:
    - echo &quot;Since build_b and test_b run slowly, this deploy job will run much later.&quot;
environment: production</pre>
<br>

== 父子流水线 ==

随着流水线变得越来越复杂，一些相关的问题开始出现：

* 分阶段结构，一个阶段中的所有步骤必须在下一阶段的第一个作业开始之前完成，这会导致等待，从而减慢速度。
* 单一全局流水线的配置变得难以管理。
* 使用 include 的导入会增加配置的复杂性，并可能导致无意中重复作业的命名空间冲突。
* 流水线 UX 有太多的作业和阶段需要处理。

此外，有时流水线的行为需要更加动态。选择启动（或不启动）子流水线的能力是一项强大的能力，尤其是在动态生成 YAML 的情况下。

<br>

父子流水线将配置分成多个文件，使事情变得更简单。 您可以将父子流水线与以下结合：

* rules 关键字：例如，仅在该区域发生更改时才触发子流水线。
* include 关键字：引入常见行为，确保您不会重复自己。
* 子流水线内部的 DAG 流水线，获取两者的所有优势。

<br>

<pre>stages:
- triggers

trigger_a:
stage: triggers
trigger:
    include: a/.gitlab-ci.yml
rules:
    - changes:
        - a/*

trigger_b:
stage: triggers
trigger:
    include: b/.gitlab-ci.yml
rules:
    - changes:
        - b/*

示例子流水线 a 配置，位于 /a/.gitlab-ci.yml 中，使用 DAG needs 关键字：

stages:
- build
- test
- deploy

image: alpine

build_a:
stage: build
script:
    - echo &quot;This job builds something.&quot;

test_a:
stage: test
needs: [build_a]
script:
    - echo &quot;This job tests something.&quot;

deploy_a:
stage: deploy
needs: [test_a]
script:
    - echo &quot;This job deploys something.&quot;
environment: production

示例子流水线 b 配置，位于 /b/.gitlab-ci.yml 中，使用 DAG needs 关键字：

stages:
- build
- test
- deploy

image: alpine

build_b:
stage: build
script:
    - echo &quot;This job builds something else.&quot;

test_b:
stage: test
needs: [build_b]
script:
    - echo &quot;This job tests something else.&quot;

deploy_b:
stage: deploy
needs: [test_b]
script:
    - echo &quot;This job deploys something else.&quot;
environment: production</pre>
也可以将作业设置为在触发子流水线之前或之后运行，例如，如果您有通用的设置步骤或最后的统一部署。