如何让Pulumi在删除AWS API Gateway方法时触发重部署？

问题根源 #

基于资源ID的触发器仅追踪现有资源的变更（新增/修改），当Method或Integration资源被删除时，对应的ID会从依赖列表中移除，但Pulumi默认的依赖追踪不会主动感知到列表元素的减少，因此不会触发Deployment更新。要解决这个问题，需要让触发器依赖所有相关资源的完整状态变化，包括资源的删除操作。

具体解决方案 #

核心思路是：将所有Method和Integration资源的唯一标识（URN）聚合为一个计算值，确保资源新增、修改、删除时该值都会发生变化，从而触发Deployment重部署。

代码示例（静态资源场景） #

如果是预先定义的固定Method和Integration：

import * as pulumi from "@pulumi/pulumi";
import * as aws from "@pulumi/aws";

// 已创建的API Gateway实例
const restApi = new aws.apigateway.RestApi("my-api");

// 示例Method与Integration资源
const getMethod = new aws.apigateway.Method("get-root", {
  restApi: restApi,
  resourceId: restApi.rootResourceId,
  httpMethod: "GET",
  authorization: "NONE",
});

const getIntegration = new aws.apigateway.Integration("get-root-proxy", {
  restApi: restApi,
  resourceId: restApi.rootResourceId,
  httpMethod: getMethod.httpMethod,
  type: "HTTP_PROXY",
  integrationHttpMethod: "GET",
  uri: "https://example.com",
});

// 聚合所有相关资源的URN（Pulumi全局唯一标识）
const allResourceUrns = pulumi.all([
  getMethod.urn,
  getIntegration.urn,
]);

// 生成触发器值：排序后转为字符串，避免资源顺序变化导致误触发
const redeployTrigger = allResourceUrns.apply(urns => JSON.stringify(urns.sort()));

// 配置Deployment的触发器
const deployment = new aws.apigateway.Deployment("api-deployment", {
  restApi: restApi,
  triggers: {
    redeploy: redeployTrigger,
  },
  description: pulumi.interpolate`Deployed at ${new Date().toISOString()}`,
});

代码示例（动态资源场景） #

如果是循环创建的批量Method和Integration：

import * as pulumi from "@pulumi/pulumi";
import * as aws from "@pulumi/aws";

const restApi = new aws.apigateway.RestApi("my-api");
const resourcePaths = ["/users", "/posts", "/comments"];

const methods: aws.apigateway.Method[] = [];
const integrations: aws.apigateway.Integration[] = [];

// 批量创建资源
for (const path of resourcePaths) {
  const apiResource = new aws.apigateway.Resource(`resource-${path}`, {
    restApi: restApi,
    parentId: restApi.rootResourceId,
    pathPart: path.slice(1),
  });

  const getMethod = new aws.apigateway.Method(`get-${path}`, {
    restApi: restApi,
    resourceId: apiResource.id,
    httpMethod: "GET",
    authorization: "NONE",
  });
  methods.push(getMethod);

  const getIntegration = new aws.apigateway.Integration(`proxy-${path}`, {
    restApi: restApi,
    resourceId: apiResource.id,
    httpMethod: getMethod.httpMethod,
    type: "HTTP_PROXY",
    integrationHttpMethod: "GET",
    uri: `https://example.com${path}`,
  });
  integrations.push(getIntegration);
}

// 聚合所有动态创建资源的URN
const allResourceUrns = pulumi.all([
  ...methods.map(m => m.urn),
  ...integrations.map(i => i.urn),
]);

const redeployTrigger = allResourceUrns.apply(urns => JSON.stringify(urns.sort()));

const deployment = new aws.apigateway.Deployment("api-deployment", {
  restApi: restApi,
  triggers: {
    redeploy: redeployTrigger,
  },
});

关键说明 #

1. 使用URN而非ID：URN是Pulumi全局唯一的资源标识，比AWS原生的资源ID更可靠，能确保资源的全生命周期状态被追踪。
2. 排序URN数组：避免因资源创建顺序变化导致触发器值无意义变更，仅当实际资源集合变化时才触发部署。
3. pulumi.all的作用：聚合所有资源的状态，当任意资源被删除时，聚合结果的长度或内容会变化，进而触发触发器值更新。

内容的提问来源于stack exchange，提问作者simon.ro