https://docs.google.com/forms/d/e/1FAIpQLSejhdh-0x2C0OHdVz9xabGYww3PWtOOZ1MwNLARZIt3s5ioYg/viewform

https://docs.google.com/forms/d/e/1FAIpQLSejhdh-0x2C0OHdVz9xabGYww3PWtOOZ1MwNLARZIt3s5ioYg/viewform
they have a google forums just for secrets store linked on the blog, would be a better way of giving feedback for it
https://blog.cloudflare.com/secrets-store-beta/

Google Docs

Cloudflare Secrets Store beta: provide your feedback

Thank you for trying out our beta version of our Secrets Store offering. Your feedback helps us improve the product. This survey should take about 5 minutes.

CChaika https://docs.google.com/forms/d/e/1FAIpQLSejhdh-0x2C0OHdVz9xabGYww3PWtOOZ1MwNLAR...

st3fan•4/19/25, 3:04 AM

I filled in this form with some thoughts

mr.niko.la•4/19/25, 6:44 AM

"I have a Cloudflare domain with a CF Pages frontend that connects to my backend via a Cloudflare tunnel. However, my CORS settings are being bypassed.

I've tried the following without success:

No Pages rules or transform rules are active
Purged cache completely
Set development mode on Pages to bypass caching

Despite these efforts, requests are still only being allowed from a specific subdomain origin.

Any suggestions on how to properly configure CORS for my setup?"

Or how to disable cf pages Cors and let my backend handle it ?

st3fan•4/19/25, 1:22 PM

Is there some guidance on how to split up or combine endpoints in workers? I have a small API and a webhook - is it common to combine it all in a single worker? Or is it better to split it up in multiple? Sorry I know this is a vague question. New to serverless design patterns.

Sst3fan Is there some guidance on how to split up or combine endpoints in workers? I hav...

st3fan•4/19/25, 1:35 PM

ChatGPT actually gave some good tips ...

When to keep multiple endpoints in a single worker:
• You want shared logic, like auth, middleware, routing, or common utility functions.
• You want faster cold starts (Cloudflare Workers are fast, but more workers = more potential cold starts).
• You’re trying to simplify deployment, versioning, and observability.
• Your app is a cohesive unit, like a backend API or a website.

When to split into separate workers:
• You have very different workloads (e.g., image resizing vs. data API).
• You want to enforce stricter limits or security boundaries (e.g., public vs internal APIs).
• You’re deploying to different teams or lifecycles (e.g., teams own different services).
• You hit size/dependency limits (1MB per Worker code bundle).
• You want to deploy independently and reduce blast radius of changes.

plgingras88•4/19/25, 3:21 PM

Following the announcement of the beta worker container, it seems like we'll be able to select an instance type, with vcpu, etc. Do we have an idea of what these instance types may look like ? Will these be based on AWS EC2, etc. Otherly said, can we expect some power, like 4x vCPU of 3-4Ghz with 4-8Gb or ram ?

Sst3fan ChatGPT actually gave some good tips ... ✅ When to keep multiple endpoints in a...

Isaac McFadyen•4/19/25, 4:57 PM

FYI it's wrong about the size limit, that was changed to 10MB now for paid plans. 3MB on free though.

IIsaac McFadyen FYI it's wrong about the size limit, that was changed to 10MB now for paid plans...

st3fan•4/19/25, 5:10 PM

Maybe this is a bug in the UI then? I can’t show now but it clearly said “validation failed max size is 1024” for me

ItsWendell•4/19/25, 5:20 PM

I keep hitting 524 when I work with the Google Vertex Gemini 2.5 Flash API's, it's doing some complex thinking work and function / tool calling to process documents into a different format in multiple steps. I use workflows do handle this. The problem is that even though I use streaming, the tool call itself / thinking on the gemini-side might take over 100 seconds occasionally. That will trigger a 524 error because of Cloudflare's connection limits if I understand correctly. I already use SSE APIs, but since I force a complex function call for the AI to fill in as it's initial step, sometimes no bytes are sent within those first 100 seconds.

https://developers.cloudflare.com/fundamentals/reference/connection-limits/#configurable-limits

But it looks like this is only configurable through Enterprise accounts, are there any clever work-arounds?

Sst3fan Maybe this is a bug in the UI then? I can’t show now but it clearly said “valida...

st3fan•4/19/25, 5:55 PM

I'll try via the API (since i can't set secrets with newlines in it through wrangler)

Sst3fan I'll try via the API (since i can't set secrets with newlines in it through wran...

Isaac McFadyen•4/19/25, 5:57 PM

This is for a secret, not a Worker. You discussed splitting Workers above because of the size limit but they're completely different products

Isaac McFadyen•4/19/25, 5:57 PM

The Worker itself can be 3MB on Free and 10MB on Paid plans.

Isaac McFadyen•4/19/25, 5:57 PM

The secrets in the Worker can be max 1KB (secret = encrypted env var, basically)

st3fan•4/19/25, 6:22 PM

Ah ok I misunderstood. The worker size was never a problem for me. I don'tt hink I said anything about that.

st3fan•4/19/25, 6:23 PM

oh sorry i see the confusion .. no problem

st3fan•4/19/25, 7:28 PM

Can I use global variables in a worker? Will that survive during invocations? Cheap cache.

Sst3fan Can I use global variables in a worker? Will that survive during invocations? Ch...

Ashley•4/19/25, 7:42 PM

No, in short. Invocations are rapidly killed, it’s not really like Lambda where it may stick around for a while.

zegevlier•4/19/25, 7:46 PM

Alternatively, you can use the (free) cache API: https://developers.cloudflare.com/workers/runtime-apis/cache/

st3fan•4/19/25, 7:48 PM

Ah I was thinking of keeping a Hono router around. To save CPU time for setup.

Sst3fan Ah I was thinking of keeping a Hono router around. To save CPU time for setup.

Isaac McFadyen•4/19/25, 7:49 PM

The way hono is designed, you keep the router in the global scope anyway - it's only initialized once at startup.

Isaac McFadyen•4/19/25, 7:50 PM

i.e. https://hono.dev/docs/getting-started/cloudflare-workers

const app = new Hono()

export default app;

const app = new Hono()

export default app;

st3fan•4/19/25, 7:50 PM

Yes but if I understand correctly it would be initialized for every http request handled?

Sst3fan Yes but if I understand correctly it would be initialized for every http request...

Isaac McFadyen•4/19/25, 7:50 PM

Nope, because anything outside of the

fetch

fetch

handler (which Hono handles for you, in that example) is "global"

st3fan•4/19/25, 7:51 PM

That contradicts what @Ashley said earlier?

st3fan•4/19/25, 7:54 PM

I can add some logging to find out

Ashley•4/19/25, 7:58 PM

I’ve not used Hono tons, but I figured you wanted to cache something that was mutable which definitely won’t work typically.

Ashley•4/19/25, 7:58 PM

A given isolate has its own scope, but isolates are not necessarily long-lived. An isolate may be spun down and evicted for a number of reasons
Because of this, it is generally advised that you not store mutable state in your global scope unless you have accounted for this contingen

Ashley•4/19/25, 7:58 PM

From https://developers.cloudflare.com/workers/reference/how-workers-works/

Cloudflare Docs

How Workers works · Cloudflare Workers docs

The difference between the Workers runtime versus traditional browsers and Node.js.

st3fan•4/19/25, 8:08 PM

Hono, a GitHub api client, a webhook router. I think j it is all stateless. Thanks that is a good document!

st3fan•4/19/25, 8:10 PM

Yeah sounds like no guarantees but may shave some milliseconds off requests if I keep things around.

Walshy•4/19/25, 9:28 PM

Yeah if you re-use the isolate, what you put in global scope still exists. It can be a way to re-use across requests but there's no guarantee any 2 requests will hit the same isolate.

vg•4/20/25, 8:48 AM

For deploying a static Angular site, is it recommended to use Workers Static or Pages?

Walshy•4/20/25, 11:04 AM

?pages

SuperHelpflare•4/20/25, 11:04 AM

As announced during Developer Week, Cloudflare recommends all new projects use Workers Assets. Cloudflare Pages will continue to be supported, but, going forward, all of our investment, optimizations, and feature work will be dedicated to improving Workers. If there is something missing that you think Workers needs, please send it in #workers-and-pages-feature-suggestions

st3fan•4/20/25, 3:30 PM

I have a thing running!

st3fan•4/20/25, 3:32 PM

But I have two "Uncaught exception" errors. I found those back in the logs and they are both "The script will never generate a response." errors with a http status 500. But there are no stack traces anywhere. What should I do to see those? (Assuming the platform catches them - I don't have any exception handlers because I wanted to see it from the platform side first)

st3fan•4/20/25, 3:42 PM

wrangler tail

wrangler tail

also tells me Exception ThrownException Thrown but no details other than Error: The script will never generate a response.Error: The script will never generate a response.

st3fan•4/20/25, 3:42 PM

Or is that the exception

st3fan•4/20/25, 7:36 PM

I was complaining about the secrets manager not accepting a PKCS#8 private key earlier but today I found out that there is also wrangler secretwrangler secret to manage a different set of secrets ... which takes longer secret values just fine. Competing technologies?

Sst3fan But I have two "Uncaught exception" errors. I found those back in the logs and t...

ChaikaOP•4/20/25, 7:43 PM

generally it's because you're leaking promises/async stuff across requests in the global scope

"The script will never generate a response" happens when the runtime sees that it has finished executing all the JavaScript microtasks for a particular request, and no I/O operations have been scheduled

can be caused by a library you are using as well, there's a few that could cause it that are known issues, as well as just any that assumes globals are fine/persisting promises, which is ok in most other environments

Sst3fan I was complaining about the secrets manager not accepting a PKCS#8 private key e...

ChaikaOP•4/20/25, 7:44 PM

wrangler secrets are much older, they're per worker and part of it, Secret store is account wide

keerth•4/20/25, 7:53 PM

Failed to load module script: Expected a JavaScript module script but the server responded with a MIME type of "text/html". Strict MIME type checking is enforced for module scripts per HTML spec.

Failed to load module script: Expected a JavaScript module script but the server responded with a MIME type of "text/html". Strict MIME type checking is enforced for module scripts per HTML spec.

I keep seeing this error when deploying, it's happening for a vite react app that I've deployed, and I've edited the code and redeployed

keerth•4/20/25, 7:53 PM

not sure if anyone seen it before, but can't figure out why it's happening

keerth•4/20/25, 7:53 PM

also I'm using the same git as the template, just even changing some code and redeploying makes this error occur

keerth•4/20/25, 7:54 PM

https://damp-violet-825c.keerthanenr.workers.dev/

keerth•4/20/25, 7:54 PM

example app here

keerth•4/20/25, 7:57 PM

https://github.com/keerthanenr/damp-violet-825c

GitHub

GitHub - keerthanenr/damp-violet-825c

Contribute to keerthanenr/damp-violet-825c development by creating an account on GitHub.

keerth•4/20/25, 7:57 PM

here's the repo for it

keerth•4/20/25, 7:57 PM

if anyone can see anything that could be causing this would appreciate it!

CChaika generally it's because you're leaking promises/async stuff across requests in th...

st3fan•4/20/25, 8:00 PM

I think my code is fine. I'm starting to think it is octokit that doesn't work well on Workers.

Kkeerth if anyone can see anything that could be causing this would appreciate it!

ChaikaOP•4/20/25, 8:00 PM

bump your compat date back to 2025-03-31

https://docs.google.com/forms/d/e/1FAIpQLSejhdh-0x2C0OHdVz9xabGYww3PWtOOZ1MwNLARZIt3s5ioYg/viewform

Similar Threads