How do you install phantomjs on AWS lambda?

Question

I found a 4 year old project that's supposed to install phantomjs on lambda, but either I'm doing something wrong, or things have changed since the repo was created and it no longer works. When I clone and deploy this repository, I get this error trying to run phantomjs:

{
  "errorType": "Error",
  "errorMessage": "write EPIPE",
  "code": "EPIPE",
  "stack": [
    "Error: write EPIPE",
    "    at WriteWrap.afterWrite [as oncomplete] (net.js:779:14)"
  ],
  "errno": "EPIPE",
  "syscall": "write"
}

{
  "errorType": "Error",
  "errorMessage": "html-pdf: Received the exit code '127'\n./phantomjs_lambda/phantomjs_linux-x86_64: error while loading shared libraries: libfreetype.so.6: cannot open shared object file: No such file or directory\n",
  "stack": [
    "Error: html-pdf: Received the exit code '127'",
    "./phantomjs_lambda/phantomjs_linux-x86_64: error while loading shared libraries: libfreetype.so.6: cannot open shared object file: No such file or directory",
    "",
    "    at ChildProcess.respond (/var/task/node_modules/html-pdf/lib/pdf.js:121:31)",
    "    at ChildProcess.emit (events.js:189:13)",
    "    at Process.ChildProcess._handle.onexit (internal/child_process.js:248:12)"
  ]
}

If I use the latest linux binary instead of the one that comes with this repo, I get a different but similar error about a different .so file missing.

For context, the reason I want to install phantomjs is because I want to use the node html-pdf library in my lambda.

I am on a Mac.

As an answer, I'm not looking to get past this specific error only to discover the next one. I'm looking for an answer that gets me a working phantomjs on aws lambda. If there is another node "html to pdf"/"html to png" library that doesn't require phantomjs to work and easily runs on AWS Lambda, that would be an acceptable solution, too.

Answer 1

One solution is to include the libraries required by PhantomJS in your Lambda function .zip file. I'll break this down into several steps.

Determine which libraries to include

Your error message states that a shared library is missing. Let's find exactly what PhantomJS is looking for by installing it in a Docker container and checking its runtime dependencies. The nodejs10.x AWS Lambda Runtime uses Amazon Linux 2, but the steps are similar for earlier runtimes.

On your host system, run an Amazon Linux 2 Docker container:

$ docker run -it --rm amazonlinux:2.0.20190508

In the container, install PhantomJS and inspect its dependencies:

bash-4.2$ yum install -y bzip2 tar
bash-4.2$ curl -LO https://bitbucket.org/ariya/phantomjs/downloads/phantomjs-2.1.1-linux-x86_64.tar.bz2
bash-4.2$ tar xf phantomjs-2.1.1-linux-x86_64.tar.bz2
bash-4.2# ldd ./phantomjs-2.1.1-linux-x86_64/bin/phantomjs 
        linux-vdso.so.1 (0x00007ffdd251f000)
        libz.so.1 => /lib64/libz.so.1 (0x00007f35d0439000)
        libfontconfig.so.1 => not found
        libfreetype.so.6 => not found
        ...

We see that libfontconfig and libfreetype are missing.

Install the libraries

Next, we'll download and extract the required libraries, then copy them to the host system. Note that libfontconfig depends on libexpat, so we'll install that as well.

In the container:

bash-4.2$ yum install -y yum-utils rpmdevtools
bash-4.2$ cd /tmp
bash-4.2$ yumdownloader fontconfig.x86_64 freetype.x86_64 expat.x86_64
bash-4.2$ rpmdev-extract *.rpm

Copy the libraries into a directory:

bash-4.2$ mkdir -p /deps
bash-4.2$ cp /tmp/*/usr/lib64/* /deps

Also copy the fontconfig configuration file:

bash-4.2$ cp /tmp/*/etc/fonts/fonts.conf /deps

Next, on the host, get the Docker container ID and copy the files from the container to the host. Below, lambda-node-phantom is the directory where you cloned the https://github.com/TylerPachal/lambda-node-phantom repository:

$ cd lambda-node-phantom
$ docker ps
$ docker cp <CONTAINER_ID>:/deps/ .
$ mv deps/* . && rmdir deps

Update the Lambda

Update the PhantomJS binary in your directory to the version downloaded in the Docker container above. Make sure that index.js refers to the correct name for the latest version; in version 2.1.1 it's named phantomjs.

Next, add the following line to index.js so that fontconfig finds fonts.conf in the Lambda root directory:

process.env['FONTCONFIG_PATH'] = process.env['LAMBDA_TASK_ROOT'];

Finally, re-create the Lambda function .zip file, including the native libraries and font.conf:

$ zip -y /path/to/lambda-node-phantom-dist.zip .

The -y option stores symbolic links as links instead of the referenced file. To save space, make sure that the .git directory is not included in the .zip file.

Test the Lambda

In the AWS Lambda console, upload the new Lambda function .zip file and test the Lambda function. PhantomJS should now run without errors, and your function should return "hello from phantom!" If not, check the stderr returned by the PhantomJS child process for information.

It's likely that you'll need to modify fonts.conf and/or include additional files for fonts to be rendered properly.

Answer 2

According to your log error, it looks like libfreetype.so.6 is missing from your lambda execution runtime.

You may need to create a custom lambda layer embedding this shared library. Afterwards, you may need to update your LD_LIBRARY_PATH so it also points to the shared library's directory. You may also alternately include it in your lambda deployment package.

According to AWS official doc:

To include libraries in a layer, place them in one of the folders supported by your runtime.

All – bin (PATH), lib (LD_LIBRARY_PATH)

Link here: https://docs.aws.amazon.com/lambda/latest/dg/configuration-layers.html

Make sure you embed a libfreetype.so.6 compiled for Amazon linux (or Amazon linux 2).

Execution runtimes here: https://docs.aws.amazon.com/lambda/latest/dg/lambda-runtimes.html

Good luck !