Inconsistent Puppeteer page.goto Behavior Across Browsers on Heroku Node.js

Other Questions Puppeteer
1

When running Puppeteer on Heroku, page.goto exhibits different behaviors across browsers, causing loss of login credentials in some cases. For example:

const puppeteer = require('puppeteer');

(async () => {
  const browserInstance = await puppeteer.launch({ args: ['--disable-sandbox'] });
  const newPage = await browserInstance.newPage();
  await newPage.goto('http://example.com/auth', { waitUntil: 'networkidle0' });
  await newPage.type('#userInput', 'demoUser');
  await newPage.type('#passInput', 'demoPass');
  await newPage.click('#loginButton');
  await newPage.waitForNavigation({ waitUntil: 'networkidle0' });
  await newPage.pdf({ path: 'output.pdf', format: 'A4', printBackground: true });
  await browserInstance.close();
})();
2

hey, i had this issue too. i solved it by adding a short delay after clicking login, so the creds got saved properly. may be a race conditon, try that tweak.

3

Based on my experience, I found that when using Puppeteer on Heroku, the issue sometimes lies in the asynchronous nature of the page events and the environment differences between browsers. In one project I worked on, I had to tweak the navigation timing by not only relying on networkidle0 but also by explicitly waiting for certain DOM elements to be visible before proceeding. I refined the wait criteria by using page.waitForSelector after login and added a short delay to ensure that all tokens and credentials were properly saved. This method helped stabilize the behavior across different browser instances on Heroku.

4

In my experience, the key to resolving inconsistent behavior in Puppeteer on Heroku was to examine the entire environment setup. I found that simply adding delays wasn’t a complete fix. I experimented with extending timeouts for both page navigation and element visibility, which helped to mitigate any issues arising from network delays or resource constraints on Heroku. Furthermore, I incorporated detailed logging to pinpoint exactly where the credentials were being lost. Adjusting the launch options and verifying that all browser instances were started with identical configurations also played a significant role in stabilizing the process.