Pre-download Chromium with code during deployment to prevent slow cold-starts

[maikelvanhaaren]

First of all, thank you for the effort you're putting into this package!

We recently made a transition from using NodeJS to .NET (Linux) for our consumption-based Azure Function, for various reasons. In our previous setup with NodeJS, we utilized puppeteer, and now we're using puppeteer-sharp. I've noticed a difference in the way these two packages handle the downloading of Chromium.

In the NodeJS setup, Chromium was downloaded during the npm install process, whereas with puppeteer-sharp, it is downloaded at runtime. This disparity, combined with the nature of serverless Azure Functions, occasionally leads to errors, especially during a cold start, when Chromium hasn't been downloaded yet.

Is there a way we can address this issue? One potential solution could be bundling Chromium together with the Azure Functions code so that it doesn't need to be downloaded at runtime.

[kblok]

I wrote this post a few years ago. I hope it still works. But the solution would be using containers in azure functions.

[maikelvanhaaren]

Check, thanks for the link. That also seems like a potential solution for this by including chromium within the container.

I was experimenting today by upgrading the Azure Function from Dynamic to an enterprise plan so I can have minimal one instance warm at all times in the hope to reduce the cold-start errors. On the dynamic SKU, it's all working but on the enterprise plan with the same code, puppeteer-sharp fails at launching the browser (error while loading shared libraries: libgobject-2.0.so.0: cannot open shared object file: No such file or directory).

What do you think about an option where you have the possibility to bring-your-own-chromium (BYOC)? Something like:

await browserFetcher.SetupPredownloadedChromium(package: "../chromium-linux.zip").ConfigureAwait(false);

var chromiumExecutablePath = browserFetcher.RevisionInfo(BrowserFetcher.DefaultChromiumRevision).ExecutablePath;
var launchOptions = new LaunchOptions
{
	Headless = true,
	Args = new[] { "--no-sandbox --font-render-hinting=none" },
	ExecutablePath = chromiumExecutablePath
};

await using var browser = await Puppeteer.LaunchAsync(launchOptions).ConfigureAwait(false);

[kblok]

I think I tried that in the past, back in 2019. Maybe give it chance to see if that works?

[maikelvanhaaren]

Is there a branch open with your attempt? I'm happy to give it a try.

[kblok]

No. I discarded that attempt.

[maikelvanhaaren]

OK, check. I'm gonna give it a try and see what I can come up with so we have some more concrete things to discuss 😄