Limit chrome headless CPU and memory usage

There had been a lot of discussion going around about the unpredictable CPU and Memory Consumption by Chrome Headless sessions.

As per the discussion Building headless for minimum cpu+mem usage the CPU + Memory usage can be optimized by:

• Using either a custom proxy or C++ ProtocolHandlers you could return stub 1x1 pixel images or even block them entirely.

• Chromium Team is working on adding a programmatic control over when frames are produced. Currently headless chrome is still trying to render at 60 fps which is rather wasteful. Many pages do need a few frames (maybe 10-20 fps) to render properly (due to usage of requestAnimationFrame and animation triggers) but we expect there are a lot of CPU savings to be had here.

• MemoryInfra should help you determine which component is the biggest consumer of memory in your setup.

• An usage can be:

   $ headless_shell --remote-debugging-port=9222 --trace-startup=*,disabled-by-default-memory-infra http://www.chromium.org 

• Chromium is always going to use as much resources as are available to it. If you want to effectively limit it's utilization, you should look into using cgroups

Having said the above mentioned points here are some of the common best practices to adapt when running headless browsers in a production environment:

Fig: Volatile resource usage of Headless Chrome</sup

• Don't run a headless browser:

By all accounts, if at all possible, just don't run a headless browser. Headless browsers are un-predictable and hungry. Almost everything you can do with a browser (save for interpolating and running JavaScript) can be done with simple Linux tools. There are libraries those offer elegant Node API's for fetching data via HTTP requests and scraping if that's your end-goal.

• Don't run a headless browser when you don't need to:

There are users that attempt to keep the browser open, even when not in use, so that it's always available for connections. While this might be a good strategy to help expedite session launch, it'll only end in misery after a few hours. This is largely because browsers like to cache stuff and slowly eat more memory. Any time you're not actively using the browser, close it!

• Parallelize with browsers, not pages:

We should only run one when absolutely necessary, the next best-practice is to run only one session through each browser. While you actually might save some overhead by parallelizing work through pages, if one page crashes it can bring down the entire browser with it. That, plus each page isn't guaranteed to be totally clean (cookies and storage might bleed-through).

• page.waitForNavigation:

One of the most common issues observed are the actions that trigger a pageload, and the sudden loss of your scripts execution. This is because actions that trigger a pageload can often cause subsequent work to get swallowed. In order to get around this issue, you will generally have to invoke the page-loading-action and immediately wait for the next pageload.

• Use docker to contain it all:

Chrome takes a lot of dependencies to get running properly. Even after all of that's complete, then there's things like fonts and phantom-processes you have to worry about, so it's ideal to use some sort of container to contain it. Docker is almost custom-built for this task as you can limit the amount resources available and sandbox it. Create your own Dockerfile yourself.

And to avoid running into zombie processes (which commonly happen with Chrome), you'll want to use something like dumb-init to properly start-up.

• Two different runtimes:

There can be two JavaScript runtimes going on (Node and the browser). This is great for the purposes of shareability, but it comes at the cost of confusion since some page methods will require you to explicitly pass in references (versus doing so with closures or hoisting).

As an example, while using page.evaluate deep down in the bowels of the protocol, this literally stringifies the function and passes it into Chrome, so things like closures and hoisting won't work at all. If you need to pass some references or values into an evaluate call, simply append them as arguments which get properly handled.

Reference: Observations running 2 million headless sessions