The only limit that’s specifically 1,000 is the amount of in-house subrequests per invocation, which

The only limit that’s specifically 1,000 is the amount of in-house subrequests per invocation, which afaik should just be a rejected promise on the subrequest rather than a resources exceeded warning. That warning should exclusively be CPU or memory.

Kkian The only limit that’s specifically 1,000 is the amount of in-house subrequests p...

kika•3/17/23, 2:57 PM

maybe my (or Rust's worker runtime's) error handling is leaking memory. dunno how to test for that, needless to say that the worker works okay locally. strange though, this code is WIP and littered with unwrap()unwrap() s which turn every error into a panic!()panic!()

kianOP•3/17/23, 3:08 PM

Do you see it in the headers log?

kianOP•3/17/23, 3:18 PM

How are you making the request to the Worker? It doesn’t sound like that header is getting there at all

kianOP•3/17/23, 3:20 PM

Lots of your fetch options seem un-necessary - but I’m pretty sure no-cors doesn’t allow you to send that header.

Wwarpandas ``` Access to fetch at 'https:// test-worker.sensay.workers.dev/' from origin 'h...

Chaika•3/18/23, 7:26 PM

It sounds like you are trying to fetch your Worker Script from your browser to add to a queue, on a webpage at http://localhost:8080

The point of CORS is to prevent other websites/origins from fetching your origin, preventing scripts from accessing resources that they shouldn't have access to.

"no-cors" isn't a magical escape hatch to get around this, it's a very restrictive mode to make your request safe. Stripping most headers from the request, restricting content-type, restricting http request type, and making you unable to read the response body/status/headers: https://evertpot.com/no-cors/

The answer is to change your worker to respond with the required CORS Headers, for example:
Access-Control-Allow-Origin: http://localhost:8080
Your real website won't be on localhost:8080, for local testing you could use a proxy that adds the CORS headers for you. For production use though, if you want to continue to have your script on a different website/origin then your website, you will need to add the required CORS Headers.
More info: https://developer.mozilla.org/en-US/docs/Web/HTTP/CORS
This has nothing to do with Queues or even Workers themselves, just do to with Cross-origin requests, so if you have any more questions/etc, please use #coding-help or another channel that fits better

Matt Silverlock•3/20/23, 11:48 PM

Queue consumer concurrency is going to land in just under two weeks: https://developers.cloudflare.com/queues/changelog/#consumer-concurrency-upcoming

If you have a use-case where you want to limit concurrency, you’ll want to cap it via the dashboard or wrangler before we enable autoscaling.

Changelog · Cloudflare Queues

Queue consumers will soon automatically scale up concurrently as a queues’ backlog grows in order to keep overall message processing latency down. …

MMatt Silverlock Queue consumer concurrency is going to land in just under two weeks: https://dev...

Erisa•3/21/23, 8:18 AM

@Queues

zegevlier•3/21/23, 1:06 PM

Yup, that's totally fine

MMoretti Thinking about this use case: fetch handler receives request, process and needs...

Matt Silverlock•3/21/23, 4:30 PM

Are you trying to serve a real-time response back to a user? If so, how do you intend to do that?

Matt Silverlock•3/21/23, 4:31 PM

Queues is best at async work. If you're relying on returning a live response to a client, then it may not make sense, but you'll also need to manage your upstream API (in this case, OpenAI's) rate limits as well vs. your own user demand.

Kai•3/21/23, 5:27 PM

doing something like fetch -> upgrade WS -> push WS to DO & enqueue work item w/ DO Id -> ... queue ... -> eventual return via DO WS -> close WS works decently imo. Can reuse the WS / DO as needed & is decently efficient. It'll be a bit more expensive depending on how you implement it (DOs with WS are never evicted so you pay) but it's possible to do & even implement timeouts and that sort of thing.

MMoretti In our case I want to close the connection with the user and then reply async wi...

Matt Silverlock•3/21/23, 6:44 PM

How do you get that response back to the user? e.g. what medium?

MMatt Silverlock How do you get that response back to the user? e.g. what medium?

Matt Silverlock•3/21/23, 6:44 PM

(does the user/client poll/retry?)

MMoretti I opened a support ticket yesterday. I can't enable queues on my enterprise acco...

Matt Silverlock•3/22/23, 7:56 PM

Can you DM me your account ID / reach out to your account team? Why can’t you enable it? (what’s the error?)

MMoretti I opened a support ticket yesterday. I can't enable queues on my enterprise acco...

Matt Silverlock•3/22/23, 7:56 PM

Also: queues

Unsmart•3/22/23, 10:39 PM

What happens when you press agree?

Matt Silverlock•3/22/23, 11:01 PM

(We addressed this via DMs. They’re on a legacy subscription with ~14 accounts. Addressing it directly)

KKlowner ah ha, limited to one active consumer!

Emilio•3/23/23, 6:38 PM

Bingo! Our Queues can only have 1 consumer at a time.

Jacob•3/24/23, 2:43 AM

When using ack() on each messages, under what conditions will it retry the non ack’d messages? Is it only if the worker crashes or retry() is called?

Jacob•3/24/23, 2:58 AM

From reading the change log, it looks like non ack’d messages will only be retried when the consumer crashes (throws an unhandled error.)
I’ll add some retry()’s to my consumer that try/catches everything

Hello, I’m Allie!•3/24/23, 2:39 PM

What’s the average time from submitting a task to it being processed? Assuming nothing else is currently being processed

HHello, I’m Allie!What’s the average time from submitting a task to it being processed? Assuming n...

Emilio•3/24/23, 3:40 PM

What do you mean by task? A message? As for processing, do you mean how long it takes the Queue broker to ingest the message? If so, that's done as soon as you send the message, plus whatever latency between the producer worker and the queue broker.

EEmilio What do you mean by task? A message? As for processing, do you mean how long it ...

Hello, I’m Allie!•3/24/23, 3:42 PM

Sorry, haven't touched Queues as much, so not 100% up on the terms. I meant, once a producer creates a task, how long for it to be piped into a consumer?

HHello, I’m Allie!Sorry, haven't touched Queues as much, so not 100% up on the terms. I meant, onc...

Emilio•3/24/23, 3:44 PM

That depends on a lot of variables! How large is your backlog? This will lead to a longer consumer delay. In the case where your Queue has low throughput, and you have a configured a long max_wait_timemax_wait_time then it might take the length of that delay.

EEmilio That depends on a lot of variables! How large is your backlog? This will lead to...

Hello, I’m Allie!•3/24/23, 3:45 PM

Assuming 0 backlog, is it started immediately, or does it wait a little while for more tasks to arrive?

Hello, I’m Allie!•3/24/23, 3:48 PM

By max_wait_timemax_wait_time, do you mean max_batch_timeoutmax_batch_timeout?

HHello, I’m Allie!Assuming 0 backlog, is it started immediately, or does it wait a little while fo...

Emilio•3/24/23, 3:48 PM

Yeah, in that case it will depend on what you configured for max_wait_time_msmax_wait_time_ms -- there is also some latency, which will vary based on how far away the producer worker is from the queue broker (they are not guaranteed to run on the same colo)

HHello, I’m Allie!By `max_wait_time`, do you mean `max_batch_timeout`?

Emilio•3/24/23, 3:49 PM

https://developers.cloudflare.com/api/operations/queue-create-queue-consumer
max_wait_time_msmax_wait_time_ms is a setting you configure on a consumer

Cloudflare API Documentation

Interact with Cloudflare's products and services via the Cloudflare API

The only limit that’s specifically 1,000 is the amount of in-house subrequests per invocation, which

Similar Threads

Similar Threads

Similar Threads