启用 Google Cloud Run 分块编码

Question

I have tried returning 'transfer-encoding:chunked' header to a large download but the response terminated at "31.9M" which is very close to the documented "32MB" limit.

The Unofficial FAQ states chunked encoding is now possible, but I can't seem to get it working.

Do I have to flip any flags (eg https/2) to enable streaming? Is it only possible in some regions? (I am using europe-west1)

Answer 1

The following minimal case does actually stream 45MB over Cloud Run, without any special configuration, confirmed in us-central1 and europe-west1

FROM node:14-slim
COPY index.js index.js
CMD [ "node", "index.js" ]

const http = require('http');
http.createServer((request, response) => {
  response.setHeader('Transfer-Encoding', 'chunked');

  // 45MB
  var i = 1000000;
  function nextChunk() {
    return new Promise(resolve => {
      if (i-- > 0) {
        response.write(
          '123456781234567812345678123456781234567812345678',
          () => nextChunk()
        );
      } else {
        response.end();
      }
    })
  };
  Promise.all(nextChunk())

}).listen(process.env.PORT || 8080);

Answer 2

Even after straming into chunks I'm always getting error after 32 MB reponse size, I can't get the response header 'Transfer-Encoding' = 'chunked' from Cloud Run service response:

I'm using NestJs, and I can get it in the response Headers when I execute the app locally.

I'm specifying it as:

  @Get('streamable')
  @Header('Transfer-Encoding', 'chunked')
  async streamable(@Res() res) { etc.. res.write(chunk)}