IAlso can't wait for this, a date that was mentioned in one if the github commits related to this was 15th, which is tomorrow 
SDoes anyone know of any kind of example on how to deploy PR environments with slug urls similar to how pages does it? using Pages, I can
wrangler pages deploy --branch=pr123 and it will automatically wire the preview environment with URL pr123.mypage.pages.dev. I'm struggling to do the same with a worker, because the environment and routes have to be specified in wrangler.toml. I'd be fine with deploying to the same preview env (for the variables), but then I'd need the route with the slug (pr123) to be updated to always point to the correct deployment ID. There will be multiple PRs active at the same time. Thanks!
ZHey, trying to optimise a worker that I have to use less memory (as we are seeing out of memory errors) by streaming responses from KV instead of pulling the entire 3MB value into memory. Seeing some huge performance issues after making these changes though, requests taking 25 seconds in our test environment, but locally this all run quickly. I was wondering how cloudflare workers scale when they get hit with a lot of concurrent requests to a single edge region. Also, there are limits on 6 external requests at once, including to R2 and KV, is that per request or per worker?
HPer request. Workers scale(basically) infinitely across multiple devices/requests, but requests from the same device/client may share a single isolate
ZI see, that could explain the issues we're seeing then. Why may they share the same isolate when from the same device/client?
ZDebugging this is proving a huge pain, especially as I can't get performance metrics for the CPU based part of the functions
HConnection Affinity. Basically, when you make multiple requests to a Cloudflare-protected website, your browser(or other client) may store information about the connection(which server you are connected to), so that you continue to connect to the same server over multiple requests. If you make multiple requests to the same machine for a Worker then they share an isolate(unless it is overloaded, iirc)
HSorry if this sounds confusing, I'm not sure I've worded it as clearly as can be
ZI see. Have you got any suggestions on how to understand what's causing the performance problems we're seeing? I'm basically stuck between a rock and a hard place right now. Our currently implementation uses too much memory, but the new implementation causes this strange behaviour
HDevTools might be of use here? https://developers.cloudflare.com/workers/testing/debugging-tools/#chrome-devtools
Cloudflare Docs
Debug your local and deployed Workers using a variety of tools.
ZWe basically store 3MB "lookup tables" in KV, previously we pulled these into memory, but with high concurrent request count that causes out of memory issues. So we are trying to stream the lookup table and process them per entry so we only use 64KiB per request (for the buffer)
ZBut while doing that I am seeing super weird behaviour like the R2 requests taking 10 seconds or even 25 seconds alone sometimes
ZWhich is where my theories about contention came in
ZI'll try to get that debugging working remotely
ZSorry, one more thing, the really odd behaviour is this
ZThe requests all get stuck for a while, and then basically all come back at the same instant
Z
ZThat's the call to R2
BReadableStreamBYOBReader.readAtLeast(minBytes, bufferArrayBufferView) as perhttps://developers.cloudflare.com/workers/runtime-apis/streams/readablestreambyobreader/#methods says:
Returns a promise with the next available chunk of data read into a passed-in buffer. The promise will not resolve until at least minBytes have been read.
What happens if the source stream ends with less than
minBytes available? A little bit of sanity would say that it would resolve with the buffer partially filled with the available bytes, rather than rejecting, but that contradicts the absolute statement "The promise will not resolve until at least minBytes have been read". Can someone confirm and perhaps update the docs?
YHello, I have a question about Workers pricing. There is a line written in the pricing for inter-worker sub-requests that they are not charged.
https://developers.cloudflare.com/workers/platform/pricing/
Just to make sure I understand, this means that if I make a request to a worker from a client and this worker needs to call two other workers in JsRPC, this will count as a single request (but will bill me for the execution time in ms on all the workers), right? Or have I misunderstood?
https://developers.cloudflare.com/workers/platform/pricing/
Just to make sure I understand, this means that if I make a request to a worker from a client and this worker needs to call two other workers in JsRPC, this will count as a single request (but will bill me for the execution time in ms on all the workers), right? Or have I misunderstood?
Abeg someone helps me setup #How can someone fix this sending subject to discord webhook!
AHad a couple questions on the Snippets alpha. Does anyone know if 1) There are plans to add analytics / logging for when a Snippet Worker runs into a resource limit error? 2) Plans to allow for async actions, to execute after the response has been returned (like a full Worker)?
Aion kno
Aif i can get mine to work
Aill do updates for it
Asomeone come help in #How can someone fix this sending subject to discord webhook!
Acan someone help in #How can someone fix this sending subject to discord webhook!
IAs you've already been told multiple times, please don't crosspost the same question in many different channels.
Ait doesnt show inemail worker
AI'm looking at the docs on workers rules, but it's not quite clear to me how it's supposed to work - I'm trying to make it so that hitting /foo will use a handler defined in /routes/foo.ts - but have also tried having /foo.ts -is there an example around of how this might work?
IWorkers rules define when to fire a Worker, they don't control which file is run. You'll need a router within the Worker, like https://hono.dev or itty-router.
Hono is a small, simple, and ultrafast web framework for the Edges. It works on Cloudflare Workers, Fastly Compute, Deno, Bun, Vercel, Netlify, AWS Lambda, Lambda@Edge, and Node.js. Fast, but not only fast.
Aah, okay. That's kinda of what I had been doing until now, defining a route parser and then handing off to a handler class.
AWas just hoping to make my code a little more modular and have it so that the parent didnt' need to know about the children explicitly, and could just be inherited by a config. Ultimately the goal was to have routes that were never compiled/deployed in prod.
Xno one is helping me bro
Xi use cloudflare and when i first download it shows the little task thingy in the bottom right
RThis isn't allowed it seems. The API throws an error:
Route pattern should not have query parameters [code: 10022]
Cyea you can't use query params in routes
https://developers.cloudflare.com/workers/configuration/routing/routes/
Route patterns may not contain infix wildcards or query parameters. For example, neitherexample.com/*.jpgnorexample.com/?foo=*are valid route patterns.
Conly way to catch query params is wildcard like the docs quote you linked says
RThat's what I was afraid of. I was hoping to avoid doing the route matching in the worker itself since 99% of our hits would not be matching and it'd cost us CPU time for nothing.
Cthe cpu time should be essentially zero if you check the path and don't execute whatever it is you're doing, inside of the worker itself
Crequest costs though
RYeah there's an API on the zone so I was really hoping to avoid that.
RAll because marketing wants their ?utm parameters 
Eoh my bad :(
They guys im a complete workers noob but have got everything up and running and deployed a few tests and im getting my feet wet, I get that workers aren't going to replace all server functions and you're limited to packages etc, would there be any way by the grace of jebus to be able to extract text from pdf with workers, rust/python/ts anything im open, I can use a server or third party api if I have to but id love to just do it on workers, only need to do 1-3 documents at a time per run and just save the text
AIf you can write code to execute it anywhere, why wouldn't you be able to do it in a worker?
Ti don’t know because of limitations of libraries that extract text from pdfs i guess ? also due to my ignorance im pretty dumb programmer just figuring out how everything works one function at a time
Tno libs natively supported i could see
