I'm migrating an existing worker from service worker to ES module format. Not a huge deal, with one

I'm migrating an existing worker from service worker to ES module format. Not a huge deal, with one caveat. Before I was accessing environment variables as globals, now they're are in the environment object. Are people just plumbing that environment object through layers of their code or is there a more elegant way to handle things? I'm finding myself adding

env

env

env

env to a large quanity of my functions just to get access to things where I need them.

TTin Cap I'm migrating an existing worker from service worker to ES module format. Not a ...

Hello, I’m Allie!•7/6/23, 6:59 PM

You can use https://developers.cloudflare.com/workers/runtime-apis/nodejs/asynclocalstorage to plumb variables out around your Worker.

AsyncLocalStorage · Cloudflare Workers docs

Cloudflare Workers provides an implemenation of a subset of the Node.js AsyncLocalStorage API for creating in-memory stores that remain coherent …

HHello, I’m Allie!You can use https://developers.cloudflare.com/workers/runtime-apis/nodejs/asyncl...

Tin CapOP•7/6/23, 7:08 PM

I see. So I could create an asynchronous context at the top of my fetch handler, pass my environment into it, then gain access to that anywhere in the following asynch code. That's nice.

TTin Cap I see. So I could create an asynchronous context at the top of my fetch handler,...

Hello, I’m Allie!•7/6/23, 7:09 PM

Yup.

jkomyno•7/6/23, 8:01 PM

How can I know the CPU architecture and OS platform my workers are running on? In node.js I'd use either process.archprocess.arch and process.platformprocess.platform, or os.arch()os.arch() and os.machine()os.machine(). However, there are no such options for CF Workers.

At most, with node_compat = truenode_compat = true, I get os.arch() === "javascript"os.arch() === "javascript" and process.platform === browserprocess.platform === browser

Jjkomyno How can I know the CPU architecture and OS platform my workers are running on? I...

Cyb3r-Jak3•7/6/23, 8:04 PM

Why do you need to know the architecture and OS? That stuff shouldn’t matter in the code

CCyb3r-Jak3 Why do you need to know the architecture and OS? That stuff shouldn’t matter in ...

jkomyno•7/6/23, 8:05 PM

I'd like to try importing a native addon (which is platform dependent) into a Cloudflare Worker. The addon would be built with napi.rs

Jjkomyno I'd like to try importing a native addon (which is platform dependent) into a Cl...

Cyb3r-Jak3•7/6/23, 8:08 PM

I don’t believe worker code runs low enough where stuff like arch is accessible. All code runs in workerd which is V8 engine

Jjkomyno I'd like to try importing a native addon (which is platform dependent) into a Cl...

jkomyno•7/6/23, 8:08 PM

(I had the strong suspicion this wouldn't be possible and that I'd have to use WebAssembly instead. I'm just asking here to get confirmation)

Jjkomyno (I had the strong suspicion this wouldn't be possible and that I'd have to use W...

Hello, I’m Allie!•7/6/23, 8:23 PM

You don't get to run native anything on Workers. It's either WASM, or JS.

Tin CapOP•7/6/23, 8:59 PM

Is it dumb to run WebAssembly inside a JavaScript worker? I have some shared code in WebAssembly that I'm running in many runtimes, it would be nice to run that inside my existing JavaScript worker

Tin CapOP•7/6/23, 9:04 PM

I'm sure there is some overhead for invoking the WebAssembly runtime. Right now my worker takes about 2ms of CPU time. The algorithm is pretty trivial and already in my JavaScript worker. I just want to replace it with the WebAssembly version so I don't have to maintain this algorithm in multiple languages. Any idea how much overhead the WebAssembly runtime has?

TTin Cap I'm sure there is some overhead for invoking the WebAssembly runtime. Right now ...

Isaac McFadyen•7/6/23, 9:08 PM

The main reason WASM takes longer is that there is no runtime, really (I mean there is but not really). By that I mean if you write some code in Go, or Rust, etc, then the entire garbage collector or standard library has to be shipped with the WASM, which means it takes much longer to boot.

Isaac McFadyen•7/6/23, 9:08 PM

Versus JavaScript, which can be run directly on a V8 isolate, meaning scripts share the garbage collector etc (isolated in isolates but more shared than WASM)

Isaac McFadyen•7/6/23, 9:09 PM

Pretty much the only way to know is to benchmark yourself with your specific WASM code.

Tin CapOP•7/6/23, 9:12 PM

My hope was to actually invoke my WebAssembly from inside a JS worker. I'm realizing now this might not be possible? I'm not seeing WebAssembly as part of the Worker runtime. Is that right?

Tin CapOP•7/6/23, 9:13 PM

Looks like v8 supports it, but not sure about the CF version

TTin Cap My hope was to actually invoke my WebAssembly from inside a JS worker. I'm reali...

Hello, I’m Allie!•7/6/23, 9:20 PM

It does. That’s how #

workers-rs works

Tin CapOP•7/6/23, 9:22 PM

Great. I'm just getting into it. I rewrote my algorithm in AssemblyScript and compiled to WASM. I'm just about to start trying to use it in my Worker.

TTin Cap My hope was to actually invoke my WebAssembly from inside a JS worker. I'm reali...

jkomyno•7/6/23, 9:26 PM

Today I've verified that this is possible!

Jjkomyno Today I've verified that this is possible!

Tin CapOP•7/6/23, 9:27 PM

Awesome

jman•7/7/23, 2:03 AM

I noticed today that req.url seems to be defined to be relative to the routes in your wrangler config, even when you are developing locally.

This has issues when developing locally with libraries like next-auth, since they use the req.url to generate the callback urls.

I'm wondering if anyone here would know of a reason for it to be this way?

Jjman I noticed today that req.url seems to be defined to be relative to the routes in...

gonz•7/7/23, 2:27 AM

try with --host 127.0.0.1:8787--host 127.0.0.1:8787 though it won't work with --remote--remote

Ggonz try with `--host 127.0.0.1:8787` though it won't work with `--remote`

jman•7/7/23, 3:35 AM

That does seem to resolve the issue!

It would still be good to understand why/if this is actually needed if anyone had ideas.

gritgodx•7/7/23, 3:37 AM

does anyone know if cloudflare reserves short subdomains for services like workers? (also zero trust seems to have no subdomains available unless you got 3+ chars)

Shirokawa•7/7/23, 7:09 AM

Has anyone done some load testing with k6 on Workers?
Basic stuff runs without hitch, but trying to test e.g. a 1000 simultaneous users/requests gives a bunch of timeouts, though as far as I understand Workers itself should be able to handle that no problem. Makes me think that perhaps the requests are getting blocked/throttled by Cloudflare.

nuno•7/7/23, 9:02 AM

Hey guys, there's any way to do a similar deploy flow from pages to workers, like, in each commit the worker publishes/deploys itself?

Codedec1296•7/7/23, 12:16 PM

Hey, can anyone help me to connect D1 database to worker...have checked the document also but I was doing it wrong somewhere. Can you please provide sample code for this.

Codedec1296•7/7/23, 12:27 PM

As I am new to this cloudflare workers and wrangler ... don't know to integerate .

protochron•7/7/23, 1:17 PM

are there any tips for including packages in a JS worker that include wasm? It looks like the worker build is generating a wasm2.__wbindgen_start()wasm2.__wbindgen_start() line in the bundled js, but the worker claims it can't find it even though it's a function in the bundled wasm file

banbanboi•7/7/23, 4:03 PM

i am encountering a 403 error when accessing the workers endpoint through a deployed web ui, but its working in my localhost. Did I missed something

jkomyno•7/7/23, 4:45 PM

I'd like to hear a Cloudflare devs's opinion on this PR for wasm-bindgen to avoid the "JavaScript Plumbing" step currently needed in JS+Wasm workers: https://github.com/rustwasm/wasm-bindgen/pull/3510.
Hopefully the PR's description, comments, and clarifications are easy to understand

IIsaac McFadyen You could try https://github.com/robertcepa/toucan-js/tree/master/packages/touca...

Tin CapOP•7/7/23, 9:56 PM

Can confirm. I just setup toucan-js in my ES module project. Works great.

Tin CapOP•7/7/23, 9:57 PM

Ended up being pretty simple:

function logErrorToSentry(request: Request, env: Env, ctx: ExecutionContext, err: any) {
    const sentry = new Toucan({
        dsn: env.SENTRY_DSN,
        context: ctx,
        environment: env.ENVIRONMENT,
        request,
    });
    sentry.captureException(err);
}

function logErrorToSentry(request: Request, env: Env, ctx: ExecutionContext, err: any) {
    const sentry = new Toucan({
        dsn: env.SENTRY_DSN,
        context: ctx,
        environment: env.ENVIRONMENT,
        request,
    });
    sentry.captureException(err);
}

Gytis•7/7/23, 11:15 PM

is it possible to have a single worker without autoscaling?

GGytis is it possible to have a single worker without autoscaling?

Cyb3r-Jak3•7/7/23, 11:24 PM

There isn’t but why would you want that?

GGytis is it possible to have a single worker without autoscaling?

Hello, I’m Allie!•7/7/23, 11:26 PM

Depends. What are you trying to do?

CCyb3r-Jak3 There isn’t but why would you want that?

Gytis•7/7/23, 11:26 PM

I'm planning on syncing AWS infra to KV, it should be alright to have autoscaling but in some cases it might be difficult to implement it if it's multiple jobs

Gytis•7/7/23, 11:27 PM

Always an option to host adaptor myself

GGytis I'm planning on syncing AWS infra to KV, it should be alright to have autoscalin...

Hello, I’m Allie!•7/7/23, 11:27 PM

You could use a Cron trigger to sync it every minute or so

Hello, I’m Allie!•7/7/23, 11:27 PM

Or however often you want

Gytis•7/7/23, 11:28 PM

I'm not a fan of Cron. Would rather have reactive arch

GGytis I'm not a fan of Cron. Would rather have reactive arch

Hello, I’m Allie!•7/7/23, 11:28 PM

Like, something you trigger yourself?