Pages is the recommended way for now With convergence static assets will come to Workers officially

Pages is the recommended way for now

With convergence static assets will come to Workers officially (Sites was always quite a hack) but that's still a way off

rob•5/21/24, 9:39 PM

welp. lol . alright so for WFp if someone wants to deploy static frontend jsut use pages API or something liek that? it won't be able to be served from the same worker ?

WWalshy Pages is the recommended way for now With convergence static assets will come t...

Hello, I’m Allie!•5/21/24, 10:45 PM

Is there a Pages For Platforms that you can call via a Dispatcher?

Hello, I’m Allie!•5/21/24, 10:45 PM

Or something similar

WalshyOP•5/21/24, 10:47 PM

no, you can spin up a site though and invoke it (we have customers doing similar just not through wfp)

Ttanushree │ Workers PM Second argument hasn't been implemented (publicly) yet. It's a way to pass bindi...

avogadro•5/22/24, 1:04 PM

I want to pass some custom args to the user worker without modifying the original request. Is this not possible right now? could be a good use case of the second parameter in dispatcher.get(..).

WalshyOP•5/22/24, 1:05 PM

jsrpc

WalshyOP•5/22/24, 1:05 PM

You could implement your own fetch handler and pass in whatever you want

WalshyOP•5/22/24, 1:06 PM

https://developers.cloudflare.com/workers/runtime-apis/rpc

Cloudflare Docs

Remote-procedure call (RPC) · Cloudflare Workers docs

The built-in, JavaScript-native RPC system built into Workers and Durable Objects.

avogadro•5/22/24, 1:09 PM

Gotcha, I guess I can do that. Would the cfcf object still be stripped when passing the RequestRequest through?

ehrenmann•5/22/24, 7:37 PM

hi guys, im new to all this so this may be a stupid question.
I want to make a worker that gets visitors data from headers (useragent...) and want to save this somewhere. i saw that there is hyperdrive and as far as i understand it, i can connect my db. are there any limits for reading/writing and also for this kind of worker should i get worker for platforms? are there any benefits?
thanks

Abdallah Abis•5/22/24, 9:58 PM

Hey Folks,
In docs here https://developers.cloudflare.com/cloudflare-for-platforms/workers-for-platforms/get-started/dynamic-dispatch/
I see that await userWorker.fetch(request)await userWorker.fetch(request) will throw an error with message that starts with Worker not foundWorker not found if the worker is not found.
What if I'm using this with RPC, when the worker is not found I get internal error.internal error. and not Worker not foundWorker not found how can I configure this, or how would I know that if the worker threw exception or not found

Cloudflare Docs

Create a dynamic dispatch Worker · Cloudflare for Platforms docs

After you have created a dispatch namespace, you can fetch any user Workers in the namespace using a dynamic dispatch Worker. The dynamic dispatch …

Eehrenmann hi guys, im new to all this so this may be a stupid question. I want to make a ...

Owl Reddy•5/25/24, 2:07 AM

I'm the newest one here but I guess you can use cloudflare analytics for that kind of data as it's easy to do time series queries with that besides normal queries. Hyperdrive needs Postgres which you gotta deploy urself or use hosted services and it's maily for read heavy workloads.

Abdallah Abis•5/28/24, 11:12 AM

Hey folks,
Doing a PUTPUT request to /secrets/secrets returns this:

{
    "result": {
        "name": "TEST",
        "type": "secret_text"
    },
    "success": true,
    "errors": [],
    "messages": []
}

{
    "result": {
        "name": "TEST",
        "type": "secret_text"
    },
    "success": true,
    "errors": [],
    "messages": []
}

But doing a

GET

GET

just after that to the same endpoint returns this:

{
    "result": [],
    "success": true,
    "errors": [],
    "messages": []
}

{
    "result": [],
    "success": true,
    "errors": [],
    "messages": []
}

Any idea why secrets aren't being set?

AAbdallah Abis Hey folks, Doing a `PUT` request to `/secrets` returns this: ```JSON { "resu...

Abdallah Abis•5/28/24, 4:50 PM

I found the issue, but have no idea how to fix it.
So assuming x-script-namex-script-name is deployed to both namespace-devnamespace-dev and namespace-prodnamespace-prod (namespace-dev first)

If I do a PUTPUT to /workers/dispatch/namespaces/namespace-prod/scripts/x-script-name/secrets/workers/dispatch/namespaces/namespace-prod/scripts/x-script-name/secrets
The API hits /namespaces/namespace-dev/scripts/namespaces/namespace-dev/scripts and not namespace-prodnamespace-prod for some reason.
I think it hits the namespace that it was deployed to first.

I did rename the worker and deploy it to namespace-prodnamespace-prod first and add a secret to it and it worked, but deploying the same renamed worker to namespace-devnamespace-dev and trying to add a secret to the last deployed worker it actually ended up added to the namespace-prodnamespace-prod one.
In other words, the dispatch namespace in this API has no effect it just looks in all the user workers on the account and picks the first one deployed and not the one in that particular workspace.

avogadro•5/28/24, 8:15 PM

I am calling return await userWorker.handleRequest(request)return await userWorker.handleRequest(request). When i look in the logs, I see that it says "outcome": "cancelled". Any ideas?
the handleRequest implementation is just return new Response('Hello World!');return new Response('Hello World!');

Aalgads Im coming back to workers-for-platforms now that its been released GA and the pr...

Isaac McFadyen•5/28/24, 9:04 PM

So I believe you're confusing latency and CPU time. 1-2ms would be a reasonable CPU time for a simple JSON worker, and 28-32ms would be a reasonable wall time for that worker. I can't imagine any scenario where a simple JSON worker is using 28-32ms of CPU time per request, that's very high.

Isaac McFadyen•5/28/24, 9:04 PM

?cpu-time

Flare•5/28/24, 9:04 PM

CPU Time: the amount of work the CPU needs to do, measured in msms. More CPU-heavy things will increase it (for example, hashing passwords). This is not the same as the time a

fetch

fetch

request takes.
Wall Time: the actual amount of time a Worker runs for, as though you measured it with a clock or your browser DevTools

Aalgads Im talking about the execution speed of the user worker to receive the payload a...

Isaac McFadyen•5/28/24, 9:08 PM

Ah OK, I misunderstood then. Just to clarify - you are not billed for ~30ms, you are billed by CPU time.

Unsmart•5/29/24, 12:03 AM

WFP has always had some latency to fetch a script

Unsmart•5/29/24, 12:17 AM

Yes you can and should expect some latency.

Aalgads So 20-30 ms per user worker is expected and the under 5ms we experienced previou...

WalshyOP•5/29/24, 2:51 AM

20ms is definitely not normal for any non-cold request

WalshyOP•5/29/24, 2:51 AM

If you've got a url I can hit happy to investigate

Unsmart•5/29/24, 2:52 AM

The 20ms is the cold start he said occasionally it will be 0/1ms from the cache

WalshyOP•5/29/24, 3:24 AM

it's definitely possible but cold starts are not common for live apps with traffic, we can't keep worker isolates around forever if we get no traffic :p

Aalgads Oh this is interesting. If I run from Chrome, I get a quick cold start and then...

WalshyOP•5/29/24, 3:24 AM

clients won't matter, you will have just hit another colo or something

WalshyOP•5/29/24, 3:25 AM

fwiw my request was < 0ms

WalshyOP•5/29/24, 3:25 AM

to load and invoke them

WalshyOP•5/29/24, 3:27 AM

as the one who writes code for these, i can confirm the clients don't matter :p

WalshyOP•5/29/24, 3:31 AM

hit it from sjc and yeah, can confirm ~20ms for a cold read

Aalgads If I just keep hitting repeat on Bruno I would expect it to come back with 0 on ...

Chaika•5/29/24, 3:31 AM

I tested with Bruno and it looks like it has no http keepalive so each request is brand new, hitting different metal/machine and potentially different colo (in the case of lax which has multiple colos in one, would all show as lax)

WalshyOP•5/29/24, 3:32 AM

We're working to cut that time down (from this specific request, it should half)

WalshyOP•5/29/24, 3:32 AM

and we'll continue improvements there

WalshyOP•5/29/24, 3:32 AM

but again, I wouldn't index on cold reads. Hot reads are what you will actually see

CChaika I tested with Bruno and it looks like it has no http keepalive so each request i...

WalshyOP•5/29/24, 3:34 AM

makes sense, this was my suspicion

Aalgads I had this issue while doing some debugging. It appeared that if I did not cons...

avogadro•5/29/24, 11:27 PM

I did some more debugging. It turns out if I return back an object like {foo: 'bar'}{foo: 'bar'} from a method on an RpcTarget, then the call to that rpc target shows up as "cancelled". (I call a method that returns an RpcTarget/RpcStub, and then call a method on that RpcTarget class instance)
but if i return a simple string like "foo bar""foo bar" then the target shows up as "OK".

I did the above in controlled environment, with the only variable being the value being returned from the rpc method.

Aavogadro I did some more debugging. It turns out if I return back an object like `{foo: '...

Isaac McFadyen•5/30/24, 2:09 AM

This is Workers for Platforms, correct?

Pages is the recommended way for now With convergence static assets will come to Workers officially

Similar Threads

Similar Threads

Similar Threads