Cloudflare Developers•3y ago

What do you mean by "not work"

UUnsmart What do you mean by "not work"

monkev_see_monkev_do•12/8/23, 5:08 AM

It’s just not showing the URL, for example if you type in Google.com, you would expect the site to load.

monkev_see_monkev_do•12/8/23, 5:11 AM

James•12/8/23, 7:37 AM

How do you enable Durable Objects? The link is useless.

Erisa•12/8/23, 8:57 AM

Go to a Worker in your dashboard, scroll down on the overview and there should be a Durable Objects section with an enable button

kian•12/8/23, 2:05 PM

The vast majority of that isn't in workerd

kian•12/8/23, 2:06 PM

You'd see the .h.h stuff and maybe some

workerd

workerd

specific overrides, but most of it is defined in Cloudflare's internal repository.

kian•12/8/23, 2:09 PM

There's weird things like requests can have 6 concurrent subrequests, but your isolate does have an overall limit (which you can hit with enough rps but not reading response bodies).

kian•12/8/23, 2:09 PM

It leads to weird scenarios where you're reading the response body even if to discard it instantly (i.e not assigning it to a variable).

kian•12/8/23, 2:11 PM

You can see the same sort of behaviour with Service Bindings

kian•12/8/23, 2:11 PM

If you make a call to a bound service but don't read the body, the bound Worker will show CanceledCanceled for that request in tail.

kian•12/8/23, 2:12 PM

Since they're effectively held open, in the event you do consume the body later, you can hit a limit for that isolate.

kian•12/8/23, 2:13 PM

As for resources on a bound Worker, yeah it's in the same isolate so memory is the primary concern as you get a refreshed CPU limit (i.e on Bundled you get another 50ms to spend in that bound Worker).

kian•12/8/23, 2:14 PM

Exceeding resources tends to make the isolate go

but I haven't tested it with a bound Worker.

Hello, I’m Allie!•12/8/23, 2:14 PM

Is there a method for black holing a Response Body?

kian•12/8/23, 2:15 PM

You'd want to store an ID in the global scope of the calling Worker as well as the ID of that Cloudflare server (call https://cloudflare.com/cdn-cgi/tracehttps://cloudflare.com/cdn-cgi/trace and store the flfl line), call a bound Worker that exceeds resource limits & hit the Worker until you see the same fl line. If your id is different, it's a new isolate.

kian•12/8/23, 2:15 PM

The flfl part is because your Worker is a per-server thing, so you might hit two different servers in the same colo which are different isolates.

kian•12/8/23, 2:16 PM

That's the primary reason that trying to reason about memory consumption, or using global variables for state, is pretty difficult on Workers.

kian•12/8/23, 2:16 PM

Your connection might have a keep-alive to that server (i.e a browser) or you might hit a new server everytime (i.e hard refresh, spamming it w/ curl, etc)

kian•12/8/23, 2:17 PM

I've had isolates that sit idle on a given server for 22+ hours since their last request during off hours.

kian•12/8/23, 2:19 PM

I believe the isolate would hang around, just that request is cancelled.

kian•12/8/23, 2:20 PM

1101 (exception) only kills the request, 1102 (resources) kills the isolate.

kian•12/8/23, 2:23 PM

When the client disconnects, all tasks associated with that client request are canceled. Use event.waitUntil() to delay cancellation for another 30 seconds or until the promise passed to waitUntil() completes.

kian•12/8/23, 2:24 PM

It seems so

kian•12/8/23, 2:24 PM

I don't know if that means disconnects as in the response body has been fully read, or a mid-request disconnect.

Constantin•12/8/23, 11:02 PM

is there a better way to group thinks like kv_namespaces under a specific env, withoud the need to use env.... each time ?

lukas•12/8/23, 11:03 PM

Kinda confused why my DO alarm is not being triggered every 10 seconds

In one of my DO endpoints I have this

this.state.storage.setAlarm(Date.now() + 10 * 1000)

this.state.storage.setAlarm(Date.now() + 10 * 1000)

checking the CF logs, it only gets triggered once (in 10 seconds after creation)

CConstantin is there a better way to group thinks like kv_namespaces under a specific env, w...

Hello, I’m Allie!•12/8/23, 11:18 PM

There’s an easy solution to all the TOML ick:

ACS•12/9/23, 12:21 AM

Hello, Is

fetch

fetch

inside workers automatically cache POST request with custom cacheKeycacheKey ?
Or I have to clone the request and change the method to GET then use the Cache API ?

Erisa•12/9/23, 12:35 AM

You need to change the method since the cache layer the api interacts with rejects and will ignore POST requests
Example here https://developers.cloudflare.com/workers/examples/cache-post-request/

Cache POST requests · Cloudflare Workers docs

Documentation for Cloudflare Workers, a serverless execution environment that allows you to create entirely new applications or augment existing ones …

EErisa You need to change the method since the cache layer the api interacts with rejec...

ACS•12/9/23, 12:44 AM

In this example, can I use cf: { cacheKey: hash }cf: { cacheKey: hash } instead of pathname ?

AACS In this example, can I use `cf: { cacheKey: hash }` instead of pathname ?

ACS•12/9/23, 1:11 AM

I just test it, It seem that Cache API respect the cfcf in the request object.

Constantin•12/9/23, 2:38 AM

what happens if a worker is in a middle of execution and a deploy just takes place? does the process gets terminated immediately?

CConstantin what happens if a worker is in a middle of execution and a deploy just takes pla...

gonz•12/9/23, 4:09 AM

I very much doubt it... wouldn't be a good user experience!

CConstantin what happens if a worker is in a middle of execution and a deploy just takes pla...

Erisa•12/9/23, 4:27 AM

By deploy do you mean you deploying a new version of the worker or us deploying a runtime update?
For your worker deploys, nothing happens. Open connections continue to run, and complete.
When we push a runtime update, Workers are given 30 seconds to finish before terminating (documented here https://developers.cloudflare.com/workers/platform/limits/#duration)

Constantin•12/9/23, 4:29 AM

yes, i was reffering to worker deployment as part of the (wrangler deploy) cmd

Erisa•12/9/23, 4:31 AM

Gotcha, then nothing happens. In-flight requests will complete without interruption with the version of the Worker they were started with. Future requests will be served with the new version of the script.

lukas•12/9/23, 4:37 AM

ok so, setAlarm only schedules an invoke in the future one time. If you want to mimic an interval, then you'd reset the alarm within the alarmalarm function correct? Is this safe?

What do you mean by "not work"

Similar Threads