what's your account ID?

ac9902443335f46a9fc60a881a41da87 please don't hack me

You are indeed over 30s

but I think I found the issue. we need to parse HTML (I know, I know) and one of the pages we are scraping is massive. I think this lib is causing trouble:

wait that is the average?

Hello guys, I have issues with the simple Cloudflare worker that acts like a proxy to a single file on a public S3. That worker was working a few months ago and now I need it again, but it doesn't work now... AWS S3 bucket is public, the file is also public, I can access that HTML file directly through the browser (masked bucket name):

https://xxxxxxx.s3.eu-central-1.amazonaws.com/error.html

And my worker is like this

addEventListener('fetch', event => {
  event.respondWith(handleRequest(event.request))
})

async function handleRequest(request) {
  return await fetch('https://xxxxxxx.s3.eu-central-1.amazonaws.com/error.html')
}

And once I attach Cloudflare Worker to some routes, those routes returns

403 Forbidden
cloudflare

but if I open that worked directly through

xxxxx.workers.dev

domain it's working as expected (HTML from S3 bucket is returned)...

for the exceeded cpu rows that i have

i'm looking only at failures

oh for exceeded ones, okay that makes sense

If you can for your use-case, I'd recommend using HTMLRewriter. It'll be significantly more efficient as it's streamed; https://developers.cloudflare.com/workers/runtime-apis/html-rewriter/

we do use that where possible, but it isn't possible for what we are doing unfortunately. we have to load the full DOM

need to pass the DOM into here:

using linkedom was a workaround to make it work on Workers. the recommended lib, JSDom, wouldn't work and is even heavier

Ahh I see, okay yeah I can definitely see how that would consume a lot of time on large pages 😅

I wouldn't be surprised if there is an infinite loop getting triggered or something

within linkedom

yeahhh I am wondering how we could limit the HTML before passing it to linkedom without destroying it

I'd share the specific page but I don't want to crash anyone's browser. it is a cast page on IMDB

it is 1.4MB

If you run those libraries to parse it locally, how fast does it run?

Anyone has some idea? First pic - workers.dev - working Second pic - real domain - not working Third pic - source code of the worker

I'd recommend posting into #1052656806058528849

as there's already ongoing support here 🙂

Oh, okay, thanks

ohh you know what... I blamed the completely wrong library. the CPU time is being consumed after it is parsed to DOM and run through the readability library. it seems to be getting stuck when we strip the HTML using another lib...

Ah well you've narrowed it down now and can hopefully reproduce and report 😄

it would probably be faster if I load it back into linkedom and use

innerText

🤣