Eall good
EHello
Ewhere can I find the wrangler file for the project in the UI?
LI am using Hono with Cloudflare Workers, when I deploy it the first requests work but the following ones remain pending forever, I look at the logs and only the preflight (OPTIONS) arrives and the request does not continue. Does anyone know what it could be?
JIs there a way to get a kind of one-to-many queue type setup with cloudflare? say I have a "new order" event, and there are various workers that want to react to that. How can I set things up to trigger multiple workers based on an event?
Dperhaps.. but the weird thing is these builds have succeeded in the past, its as if something in our account is borked, which already happened before and @Walshy helped with that. I wonder if its a related issue...
Aim starting to wonder if theres something wrong with my worker now...
My queue consumer worker is stopped at around ~15 seconds of total duration with an outcome of
The docs say that the limit is 30 seconds of cpu time for queue consumers.
AFAIK workers isnt multithreaded, so how is it possible that I am hitting the 30 second cpu time limit in 15 total seconds? This is also with a bunch of network requests, so I doubt all of that 15 seconds is cpu time
My queue consumer worker is stopped at around ~15 seconds of total duration with an outcome of
exceededCpu.The docs say that the limit is 30 seconds of cpu time for queue consumers.
AFAIK workers isnt multithreaded, so how is it possible that I am hitting the 30 second cpu time limit in 15 total seconds? This is also with a bunch of network requests, so I doubt all of that 15 seconds is cpu time
Aoh wait the timestamps on logs arent accurate are they
Aughhhhhh debugging cpu times on cloudflare is a nightmare
Ais it even possibe to figure out what the issue is? literally nothing I do will have accurate times
Athis was all working fine a few days ago, then suddenly almost every execution is failing
Athen it was all fine earlier today, and now its back
Aonly things ive changes at the most should use less cpu time, im so confused at this point and dont know what to do
A
Anevermind, the docs are just missing a step. you need to select your workdown from the dropdown at the top of the screen (the dropdown is empty initially so its not obvious its there at all)
Ai was only able to get the profiler to work once (i tried 4 other times, all of the other times it just froze when i hit stop) and it didnt show anything interesting.
Is it possible that browser rendering is using a ton of cpu time somehow? That is skipped in dev since its not supported, so thats the only thing I can think of
Is it possible that browser rendering is using a ton of cpu time somehow? That is skipped in dev since its not supported, so thats the only thing I can think of
SI wrote a koa adapter , would you use it.:cloudflare: https://www.npmjs.com/package/koa-cloudflare-workers-adapter
npm
Latest version: 0.0.1, last published: a minute ago. Start using koa-cloudflare-workers-adapter in your project by running
npm i koa-cloudflare-workers-adapter. There are no other projects in the npm registry using koa-cloudflare-workers-adapter.
Jfacing issues with browser-rendering
Jscript was working fine previously
AI think i found the issue. For some reason, after some random point, fetch requests suddenly stop processing or something? I added some logic to cancel them after a few seconds and retry, and the new requests do the same thing
Aanyone know why that might be happening?
Athe simultaneous open connections limit comes to mind, but I didnt think I was hitting that.
then I went back and re-read the docs for that just in case, and I saw they recommended canceling bodies that I don't intend on reading to free concurrent connections. Trying that now to see if thats whats causing this somehow
then I went back and re-read the docs for that just in case, and I saw they recommended canceling bodies that I don't intend on reading to free concurrent connections. Trying that now to see if thats whats causing this somehow
BIs there a way to have local development only code that gets trimmed away in production?
Ai just set an environment variable in dev, and use things like
if(env.dev)
BI don't think that trims the code though, does it?
Ai dont think so
BESBuild has
defineAh sounds like exactly what I'm looking for, thanks.
BHmm, seems like you can also use
definewrangler.tomlnpx wrangler dev --define ...
MI have max-age set to 60 on my response, yet in the browser I sometimes see this on the response. How come?
MMaybe Chrome is interfering and setting it to 14400?
MHappens in Firefox too. Is there some additional config on Cloudflare that could be causing this?
MAhh, I see that the zone is set to 4 hours. I'll try to remove that specifically for the api subdomain
MI guess this would be the solution
AI got it!
At first canceling unused responses didnt seem to do anything, but then i realized there was a case where I was breaking a loop before canceling the body stream. Fixed that and now it appears to be working!
So basically, it was not a cpu issue like the error message said. Sounds like it was just timing out, and the deadlock prevention was not working for some reason
At first canceling unused responses didnt seem to do anything, but then i realized there was a case where I was breaking a loop before canceling the body stream. Fixed that and now it appears to be working!
So basically, it was not a cpu issue like the error message said. Sounds like it was just timing out, and the deadlock prevention was not working for some reason
Amisleading, unhelpful or just wrong error messages is probably my biggest gripe with cloudflare
Sopenai's whisper costs 0.006$ per minite
worker ai's whisper uses openai, but costs 0.0039$
Am I understanding this correctly? It's much cheaper to use worker Ai, but how can this be?
worker ai's whisper uses openai, but costs 0.0039$
Am I understanding this correctly? It's much cheaper to use worker Ai, but how can this be?
1CF's GPUs are cheaper to operate than OpenAI's I would assume
IThe confusion there comes because Workers AI Whisper doesn't use OpenAI. It uses the OpenAI model but CF doesn't pay OpenAI for that (it's open-source) so they can run it much cheaper themselves.
Ethey assumed that workers ai is using openai api
SI compared the two models.
The worker AI feels average, over a second slower, and less accurate.
Are openai whisper and worker AI whisper exactly the same model?
The worker AI feels average, over a second slower, and less accurate.
Are openai whisper and worker AI whisper exactly the same model?
ENot the same model but same technology
I^ Workers AI uses Whisper
baselarge-v3baseCorrection,
large-v2Does RPC have access to bindings like D1?
HWdym RPC? A Worker?
BYeah, workers RPC.
HSure? On the other end, it is just a normal Worker
