How to use serializeTree method in Puppeteer

You need to return the recursive call up the Promise chain:

if (!pageRes || pageRes.status !== 200) {
  console.log("Website didn't load properly.")
  return isPageAccessible(page, url);
}

You can use page.evaluate() to execute JavaScript within the Page DOM context, and return the src attribute of the largest image back to Node/Puppeteer:

const largest_image = await page.evaluate(() => {
  return [...document.getElementsByTagName('img')].sort((a, b) => b.naturalWidth * b.naturalHeight - a.naturalWidth * a.naturalHeight)[0].src;
});

console.log(largest_image);

Google is doing this in Google+ and a talented developer reverse engineered it and produced http://html2canvas.hertzen.com/ . To work in IE you'll need a canvas support library such as http://excanvas.sourceforge.net/

Update

As @Alejandro pointed out in the comment, resourceType is a function and the return value is lowercased

page.on('request', request => {
    if (request.resourceType() === 'xhr')
    // do something
});

Original answer

Puppeteer's API makes this really easy:

page.on('request', request => {
  if (request.resourceType === 'XHR')
    // do something
});

You can also intercept requests with setRequestInterception, but it's not needed in this example if you're not going to modify the requests.

There's an example of intercepting image requests that you can adapt.

resourceTypes are defined here.

Error

The error happens in this line according to your error message:

let by = await page.evaluate(() => document.getElementsByClassName('text')[0].innerText); // 게시글 작성자 Text 파싱

As you are making roughly 75k requests to a website, I could imagine the website is having protective measures to ban your bot from crawling. Alternatively, the post might just not exist that you are trying to crawl.

Fix

To fix your problem, you could change your evaluate function like this. This will return undefined (instead of throwing an error) if the elements do not exist. It also improves your code by only using one page.evaluate call.

let [by, date] = await page.evaluate(() => {
    const textNode = document.getElementsByClassName('text')[0];
    const timeNode = document.getElementsByClassName('time')[0];
    return [
        textNode && textNode.innerText,
        timeNode && timeNode.innerText,
    ];
});
if (!by || !date) {
    // by or date is undefined
    console.log(`Not working for ID: ${postNumber}`);
    await element.screenshot({ path: `error-${postNumber}.png` });
}

This will make a screenshot of the pages, where an error happens. Maybe you will see that the website changed (maybe they are showing you captchas?) or the post you are trying to crawl simply does not exist.

If the screenshot is not helping, you could also use page.content() to save the HTML in an error case and have a look at it.

Puppeteer is a Node library which provides a high-level API to control Chromium over the DevTools Protocol. The API allows you to call Chromium functionality to generate PDF. The API doesn't do anything special, just providing the layer to the functionality of Chromium. As of the above, this is not the question for Puppeteer team indeed. They already answered this @ Pdf form inputs not editable.

Now to the Chromium ... "Save as PDF" option is designed to generate PDF for printing and does not support HTML form controls mapping to PDF AcroForms elements. They are not going to implement it any time soon per answer from the project member @ Issue 1024713: save-as-pdf doesn't generate editable form inputs.

Bottom line: It is not possible with Puppeteer.

The async function is surrounded by parentheses and then followed with a trailing () so that the function is immediately-invoked. See the article about IIFE on MDN for a more detailed explanation.

This is probably done because top-level await is not supported in current versions of Node.js.

$ node
Welcome to Node.js v14.12.0.
Type ".help" for more information.
> async function foo() { return Promise.resolve(true); }
undefined
> await foo()
await foo()
^^^^^

Uncaught SyntaxError: await is only valid in async function
> 

The function argument of page.evaluate() is executed in the document (browser) context and has no access to the variables of the Node.js script. You need to transfer these variables by values:

const data = await page.evaluate(SELECTOR => document.querySelectorAll(SELECTOR), TABLE_ROW_SELECTOR);

The solution was to upgrade puppeteer which includes a Chromium update. This was a Chormium bug.