How can I trigger a cloudflare worker from a non-cloudflare backend? Can I enqueue a message from a non cloudflare worker as well?

Just send an HTTP request?

In that scenario, I would be responsible for creating my own means of authenticating the request.

But gotcha. I guess I was just wondering if there was a client library of some sort or not

Not really, because any HTTP(S) request hitting a route covered by a Worker will run said Worker

There isn’t really anything there for the library to do

At least not with just a regular Worker

If you are running a GraphQL server on Workers, then there are a few libs to choose from there

Hey folks, here's something that kept me up all night: I'm making a request to a db running on my server from within a worker. Running it locally works fine, but fails once it's deployed. I've been looking at the access logs on the server and the requests coming from the deployed worker seem off: (worker on CF):

"GET /api/health HTTP/1.1" 404 19 "-" "-" 331 "-" "-" 0ms

(worker running locally):

"GET /api/health HTTP/1.1" 200 41 "-" "-" 355 "pb1@docker" "http://172.24.0.5:8090" 1ms

I notice that all the requests initiated from the deployed worker (even those with a body etc) all have only

19 bytes

request size. It's a long shot, but what are your thoughts on that?

Are you fetching an IP and/or on a non-standard port?

If not, would need to see what the

fetch

code looks like

Just your vanilla fetch:

const x = await fetch("https://db-dev.textmait.se/api/health");
console.log("Health", x.status);

I already had the thought that it could have something to do with maybe certificates, but I guess then the request would fail in a different way?

The thing that throws me off is that my reverse proxy (traefik) is the one returning the 404, it doesn't seem to be able to route it to the right docker container. So it's either an issue with the rev. proxy (traefik), or it's an issue with the request itself (cloudflare)

Okay, just deployed a minimal example, and here the requests are fine (result and server logs). Must be something with the deployment config of the other worker.

Update: It's a Cloudflare issue with custom domains. As soon as I add

route = { pattern = "api2.textmait.se", custom_domain = true, zone_id = "xxx" }

to my wrangler config, the requests fail. Any pointers to how my config is wrong?

Is the Worker you're fetching from also on

textmait.se?

Same-zone subrequests are, uh, annoying

Yup,

db-dev.textmait.se

is an A record pointing to my server,

api2.textmait.se

is managed by CF to point to the worker.

Add

toml
# https://github.com/cloudflare/workers-sdk/issues/2659
compatibility_flags = ["no_minimal_subrequests"]

to your wrangler.toml

you can track that issue to know when you can remove the compatibility flag

Thank you! What exactly does this do? Because I'm not seeing any changes after deploying with it

No idea - it's just what the Workers team recommend in that linked issue which supposedly fixes the same-zone subrequests

Huh, okay. Tbh, never heard about same-zone subrequests until now, so I'll keep that in mind. I've at least now have our service back and running by using the .workers.dev domain, so thank you very much for pointing me into the right direction!

Historically, a Worker on

foo.com

has never been able to fetch anything else on

foo.com

Using a custom domain (rather than a route) allowed it to fetch other things on

foo.com

but then

minimal_subrequests

(default as of

2022-04-05

) broke that

Wow that's fascinating. I think I wouldn't have been able to figure that out by myself

Worst case I gotta buy a second domain for the api 😅

It really should just work - comment on that it's still broken for you and the team might have other suggestions