将 mono-repo 与 Google Cloud Functions 结合使用

Question

I have multiple nodejs functions sharing the same repository and dependencies. As advised in official GCF documentation, I use a index.js file which expose multiple functions entrypoints, like this:

/**
 * GCF Entrypoints
 */

const filenames = fs.readdirSync(__dirname);
const functions = filenames.reduce((acc, filename) => {
  const isJS = /\.js/.test(filename);
  const skipped = /index|sample/.test(filename);

  if (!isJS || skipped) return acc;
  const functionName = filename.replace('.js', '');

  return {
    ...acc,
    [functionName]: require(`./${filename}`).default,
  };
}, {});

Object.assign(exports, functions);

I then execute multiple times the gcloud functions deploy command to deploy theses functions and attach their respective PubSub topics. So the whole package is uploaded and executed through different entrypoints with the same dependencies.

Is there any drawbacks or wrong-doing, if instead, I deploy the entire package as only one GC function, with one PubSub topic, and call it with the entrypoint specified in message payload?

What about function performance, warming and scalability?

It would ease the CD process a lot.

Answer 1

There is not any bad practice with your approach, If all functions within your code use the same dependencies and if are fast/simple tasks.

Keep in mind that the memory & CPU management will be shared across all your code this means that if one function needs more memory or CPU it could hurt the performance of the other requests that are running concurrently on the same instance.

The main idea of a separate each function is to isolate the necessary dependencies for every function and the processing resources.

If your code starts growing and you add more and more dependencies you will end separating the functions to have a better performance.

In this video you can find information about how your code responsiveness could be impacted by Node.js dependecies.