[BUG] Failed to launch the browser process: error while loading shared libraries: libnss3.so: cannot open shared object file

[pmakkar93]

Environment

• chrome-aws-lambda Version: 3.0.4
• puppeteer / puppeteer-core Version: 3.0.4
• OS: Mac
• Node.js Version: 12.x
• Lambda / GCF Runtime: nodejs12.x

Expected Behavior

Chrome should have load.

Current Behavior

Error: Failed to launch the browser process!
/tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

TROUBLESHOOTING: https://github.com/puppeteer/puppeteer/blob/master/docs/troubleshooting.md
at onClose (/var/task/node_modules/puppeteer-core/lib/Launcher.js:547:20)
at Interface. (/var/task/node_modules/puppeteer-core/lib/Launcher.js:537:65)
at Interface.emit (events.js:327:22)
at Interface.EventEmitter.emit (domain.js:483:12)
at Interface.close (readline.js:416:8)
at Socket.onend (readline.js:194:10)
at Socket.emit (events.js:327:22)
at Socket.EventEmitter.emit (domain.js:483:12)
at endReadableNT (_stream_readable.js:1220:12)
at processTicksAndRejections (internal/process/task_queues.js:84:21)

Steps to Reproduce

const chromium = require('chrome-aws-lambda');
const puppeteer = require('puppeteer-core');

  async convertToPdf(request) {
    try {
      console.log('Starting puppeteer browser.');

      const browser = await chromium.puppeteer.launch({
        args: chromium.args,
        defaultViewport: chromium.defaultViewport,
        executablePath: await chromium.executablePath,
        headless: true,
        ignoreHTTPSErrors: true
      });

      console.log('Puppeteer launched: ' + !!browser);


      if (request.sections.length > 75) {
        throw new Error('Only 75 report sections are allowed');
      }

      let pdfs = await Promise.all(_.map(request.sections, async (section, $index) => {
        if (section.s3Pdf) {
          let s3Response = await s3.getObject({
            Bucket: section.s3Pdf.bucket,
            Key: section.s3Pdf.key
          }).promise();

          console.log(`PDF ${$index} retrieved successfully`);
          return s3Response.Body;
        }
        const page = await browser.newPage();

        page.setViewport(request.viewPort || {
          width: 1024,
          height: 600
        });

        section.options = section.options || {};

        console.log(`Loading section ${$index}...`);
        await page.setContent(this.fromBase64(section.html), {waitUntil: ['networkidle0']});
        console.log(`Section ${$index} Load complete. Converting html to pdf using puppeteer. Options: ${JSON.stringify(section.options)}`);

        let body = await page.pdf(_.assign({format: 'letter'}, section.options));

        console.log(`PDF ${$index} created successfully`);
        page.close();
        return body;
      }));

Previously my code was running properly on nodejs8.10 supported AWS Lambda function but after forcefully upgrade to nodejs12.x, my code starts to throw this error. I tried upgrading chrome-aws-lambda and puppeteer-core version to 2.1.1 and even the latest 5.3.1 but no luck on resolving the issue.

[alixaxel]

@pmakkar93 With Node 12 the Lambda environment needs provisioning of additional shared libraries for Chromium to work, the earliest version that would work on this environment is precisely 3.0.4 (I advise you to upgrade to the latest 5.3.x though).

I think the issue that you're facing might be caused by a warm Lambda that is still keeping the files under /tmp and thus preventing these libraries from being provisioned. Please try a brand new deployment with 3.0.4 or any version since.

---

Edit: I just noticed that you're trying this on Mac? If so please check this wiki page.

[whoisjuan]

@alixaxel Any potential workaround for this? I also have this exact same problem but since 2.0.0, Puppeteer introduced a change that broke clipping beyond the viewport and this was needed to curve a lot of issues with the screenshotting mechanism.

[jmdiego]

I'm getting this same issue with:

• node 12
• chrome-aws-lambda 5.3.1
• puppeter-core 5.3.1

[yoshiokimoto]

It's not the exactly same case but I've seen same issue to initialize the browser object, missing libnss3.so. In my case it only happens when I run the lambda with aws-sam-cli in debug mode.
Actually I don't see the issue if I run the lambda at my local with normal mode (i.e. run with sam local invoke), but in debug mode (sam local invoke -d 5858) I can see the error message Failed to launch the browser process!\n/tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory.

Actually I could notice that the environment variable AWS_EXECUTION_ENV is not defined in local lambda container if I run in debug mode, it's there in normal mode on the other hand.
And when I set that variable in template.yaml file (which contains the definition of serverless application in aws sam), then the error was gone.
My experience is only in aws sam, but hope this can be a clue to resolve someone's trouble.
My environment is:

• Mac Catalina
• chrome-aws-lambda: 5.3.1
• puppeteer-core: 5.3.1
• Node: 12.x

Also, I know there are few lines which use AWS_EXECUTION_ENV as a switch to modify env variables like https://github.com/alixaxel/chrome-aws-lambda/blob/master/source/index.js#L6.
It can be better to have more robust check routine since I've seen it's not always defined.

[Sandstedt]

I get the same error, but not related to AWS, but Linux on a Windows 10 machine using WSL 2 (Ubuntu 20.04.1 LTS) and Puppetter: 5.5.0:
ChromeHeadless stderr: /home/sandstedt/_open-source/mojs/node_modules/puppeteer/.local-chromium/linux-818858/chrome-linux/chrome: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

Solved it by installing all missing dependencies listed here: Puppeteer Troubleshooting

[citak]

This may be helpful to some, to check which libs are missing just do cd /project_folder/node_modules/puppeteer/.local-chromium/linux-version/chrome-linux/ ldd chrome | grep not make sure to change the folders to your own folders, hope this helps.

[darkphoenixff4]

I just pushed up my code to AWS Lambda, and I'm getting the exact same error. New deployment (actually never been deployed before), NodeJs12, using the shelfio layer.

Also got the same error following the instructions in the Lambda Layers option. The resulting zip file, pushed up to AWS Lambda, triggers the same error, missing libnss3.so .

[damisparks]

Here is how I fixed mine.

• First, I ran an update.

sudo apt-get update

• Move to the chromium directory.

cd /project_folder/node_modules/puppeteer/.local-chromium/[linux-579032]/[chrome-linux]

but replace the bracketed directories with the directories available in your node_modules folder. There you should find the local installation of chrome and from there you ran the next command.

• Next, run this

ldd chrome | grep not

on a Linux machine to check which dependencies are missing.
Read more here Chrome headless doesn't launch on UNIX

• Run this to install the missing libraries or packages.

sudo apt-get install libpangocairo-1.0-0 libx11-xcb1 libxcomposite1 libxcursor1 libxdamage1 libxi6 libxtst6 libnss3 libcups2 libxss1 libxrandr2 libgconf2-4 libasound2 libatk1.0-0 libgtk-3-0

sudo apt-get install -y libgbm-dev

• Run this again to check if the missing packages are installed
  ldd chrome | grep not this time nothing is shown on the terminal, so you should be good.
• In your JS code, add this

const browser = await puppeteer.launch({
    headless: true,
    args: ['--no-sandbox'] });

That is it. You are good to go. Hopefully, this helps someone. 😄

[markb-trustifi]

@damisparks I have the same problem of running Puppeteer in AWS Lambda env. Lambda cannot run sudo apt-get install and not even yum install. In addition, all volumes except /tmp are read-only so, the fix provided above is irrelevant. Is there any solution for this case?

[alixaxel]

@markb-trustifi This issue is now a bit all over the place to be able to understand it and follow up on it properly.

I suggest you create a new issue with code to repro it, if that's not possible try passing dumpio: true to launch().

Chances are, that if there's a missing shared library, it will be logged there.
If I know which library it is, I can try to provision it with this package.

[sromano88-svc]

I had this error and the fix was to set lambda memory to 512 mb.

[looeee]

sudo apt-get install libpangocairo-1.0-0 libx11-xcb1 libxcomposite1 libxcursor1 libxdamage1 libxi6 libxtst6 libnss3 libcups2 libxss1 libxrandr2 libgconf2-4 libasound2 libatk1.0-0 libgtk-3-0

This resulted in an error:

Package libgconf2-4 is not available, but is referred to by another package.
This may mean that the package is missing, has been obsoleted, or
is only available from another source

E: Package 'libgconf2-4' has no installation candidate

All the other packages installed correctly. This fails:

sudo apt-get install libgconf2-4 

However, I was able to install it by using apt instead of apt-get:

sudo apt install libgconf2-4 

Once I did that puppeteer works correctly.

[ca0v]

There was a long list of dependencies, so I tried this and it worked:

sudo apt install chromium-browser

I am using Ubuntu WSL on Windows 10.

[KevinVR]

Has anybody found a solution to this for AWS Lambda? Obviously the lib can't be installed manually to the machine.

[skilbjo]

Anecdotally, I am seeing this only on the node16 runtime. When I use the node14 runtime it works fine, and when I bump to node16 I get the

Failed to launch the browser process: error while loading shared libraries: libnss3.so

error

[tomoima525]

Same here. I was trying to attempt upgrading lambda version after this update aws/aws-lambda-base-images#14 but I got the error above. It was working with v14 with the workaround mentioned in #268

[tomoima525]

@skilbjo FWIW I was able to fix the issue by using the forked package by @Sparticuz in #264

[skilbjo]

hey @tomoima525 , thanks for the heads up!

my workflow:
node: 16
lib: https://www.npmjs.com/package/chrome-aws-lambda
layer: arn:aws:lambda:us-east-1:764866452798:layer:chrome-aws-lambda:25 (from https://github.com/shelfio/chrome-aws-lambda-layer)

are you saying this works?
node: 16
lib: https://www.npmjs.com/package/@sparticuz/chrome-aws-lambda
layer: https://github.com/Sparticuz/chrome-aws-lambda/blob/13.5_bin/chrome_aws_lambda.zip?rgh-link-date=2022-05-16T16%3A59%3A35Z (from #264)

?

[Sparticuz]

That layer is currently out of date. You can build your own up to date layer using my fork though

[johntravolta1]

I didn't noticed that my serverless.yml was on runtime: nodejs16.x. When I changed it to nodejs14.x I was able to run withou this error on aws lambda.
Error: "/tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory"
Fix: Changed serverless.yml to
provider:
name: aws
runtime: nodejs14.x
and redeployed

[Sparticuz]

@johntravolta1 This repo's version only supports node14, not node16. You can support node16 by running the patch found in #274

[pencilcheck]

version 6.0.0 + node14 worked for me and under limit

[hkmsadek]

same issue. anyone solved?

[vjrngn]

I faced the same issue on AWS Lambda using Node 16.x. I did the following:

1. Update memory to 512mb. (Tweak this based on your context. I was able to get away with 256mb)
2. Use Node 14.x runtime.

Hope this helps!

[benjaminwhire]

Not really - we need Node 16.x for CDK.

[Palaksialka]

Error: Failed to launch the browser process!
/tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

node version - 16x
"puppeteer-core": "^10.4.0",
"chrome-aws-lambda": "^10.1.0"

Facing the issue.

[JoaoPedroCamargo]

Error: Failed to launch the browser process! /tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

node version - 16x "puppeteer-core": "^10.4.0", "chrome-aws-lambda": "^10.1.0"

Facing the issue.

I'm facing the same issue here. Anyone got it working in 2024?

I'm also running it in google cloud functions, is there any better way to run this script?

[Sparticuz]

I'm facing the same issue here. Anyone got it working in 2024?

Check out my fork @Sparticuz/chromium

[kisglaci]

@Sparticuz I've used @Sparticuz/chromium I've tried using it with AWS lambda and node 16.x, also using lambda layers, but it doesn't work (it does locally with public.ecr.aws/lambda/nodejs:18 docker)
tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

[Sparticuz]

@kisglaci

tmp/chromium: error while loading shared libraries: libnss3.so: cannot open shared object file: No such file or directory

That means that the AL tar isn't being extracted, which means that the environment you are using doesn't have the standard AWS environment variables

[kisglaci]

@Sparticuz Thanks for the info, I've tried using AWS lambda layers also, the AL tar is in the zip file, do you see any reason why isn't being extracted, I'm executing it in AWS lambda, so it has to have the necessary envs I think? Added an issue also: Sparticuz/chromium#285

On my local machine with public.ecr.aws/lambda/nodejs:18 docker image it does work as expected

[hex406]

Anyone facing this issue within AWS Lambda, Here's how you can fix the issue :

Just check the pull request here made by @Sparticuz :
https://github.com/alixaxel/chrome-aws-lambda/pull/274/files

You have to basically edit this file in your node_modules directory :

node_modules/chrome-aws-lambda/build/index.js

Search for this line :

    if (/^AWS_Lambda_nodejs(?:10|12|14)[.]x$/.test(process.env.AWS_EXECUTION_ENV) === true) {

and replace this line with your nodejs version in OR regex block like this : (example runtime 16 taken)

    if (/^AWS_Lambda_nodejs(?:10|12|14|16)[.]x$/.test(process.env.AWS_EXECUTION_ENV) === true) {

There will be 2 places of the same line you need to replace. Then just upload your package to AWS Lambda and you will be good to go.

Hope this helps !

[mnai01]

Downgrading to lambda Runtime: Node.js 14.x and memory 512mb worked for me

[MOSTAFAEZZAT]

The issue still exist btw I'm thinking of switching to EC2 instead of Lambda