Cloudflare Developers•3y ago

just to be accurate, KV reads are $0.50/M, and writes/lists/deletes are $5/M

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:18 PM

I.e. at same interval as the KV population?

DDaniFoldi just to be accurate, KV reads are $0.50/M, and writes/lists/deletes are $5/M

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:18 PM

Thanks I thought it was $1 so cheaper than I estimated

DaniFoldiOP•12/17/23, 7:18 PM

exactly, and storage is also $0.5/GB/mo

Original message was deleted

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:19 PM

What is cache API?

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:19 PM

Do you mean by setting cacheTtl?

DDaniFoldi exactly, and storage is also $0.5/GB/mo

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:20 PM

So best bet is write to analytics engine, then have a worker run every 300 seconds or so and pull the data out to populate the KV pair with the latest total sum requests

Original message was deleted

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:21 PM

So use cache engine to serve the data instead of KV?

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:21 PM

Ahh got it

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:21 PM

So read from Kv cache it via worker then re-read after x time period

KKDX_MW | Kaddex.xyz / Ecko.xyz So best bet is write to analytics engine, then have a worker run every 300 secon...

DaniFoldiOP•12/17/23, 7:22 PM

cron for each customer doesn't work though, you still have the 1000 subrequest limit, so unless you call the worker with a service binding, I recommend looking at DO alarms

KKDX_MW | Kaddex.xyz / Ecko.xyz So read from Kv cache it via worker then re-read after x time period

DaniFoldiOP•12/17/23, 7:23 PM

although not recommended for all workloads, you can put cache in front of KV, so you only even query KV if your cache (which is free) has expired or evicted

DDaniFoldi cron for _each customer_ doesn't work though, you still have the 1000 subrequest...

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:23 PM

Are you able to get a service binding into analytics engine?

DaniFoldiOP•12/17/23, 7:23 PM

whereas you pay for cached KV reads

DDaniFoldi although not recommended for all workloads, you can put cache in front of KV, so...

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:23 PM

This is great idea I’ll check that out, cache is per data Centre in each location?

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:23 PM

So like Sydney there are five data centers it looks like

KKDX_MW | Kaddex.xyz / Ecko.xyz Are you able to get a service binding into analytics engine?

DaniFoldiOP•12/17/23, 7:24 PM

service bindings are bindings for workers from another worker (or the same one)

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:24 PM

So there’s be five caches essentially

DaniFoldiOP•12/17/23, 7:25 PM

here's an unofficial map of all pops: https://statusmap.cloudflare.community, although there are more datacenters, not sure where you can get a list of those, but do you need a list?

DDaniFoldi here's an unofficial map of all pops: https://statusmap.cloudflare.community, al...

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:25 PM

No just trying to wrap my mind around the scale don’t need a list

DDaniFoldi cron for _each customer_ doesn't work though, you still have the 1000 subrequest...

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:26 PM

When you say call the worker with a service binding does that bypass the sub request limit?

KKDX_MW | Kaddex.xyz / Ecko.xyz When you say call the worker with a service binding does that bypass the sub req...

DaniFoldiOP•12/17/23, 7:26 PM

kind of, but there's also a limit to how many times you can call the worker

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:26 PM

So let’s say I have worked A which goes through my 4000 customers using the analytics engine and populates the KV pair. If I use a service binding from worker B to worker A and have the from on worker B would that bypass the limit?

DaniFoldiOP•12/17/23, 7:26 PM

it resets the 1000 limit, but there is also a 32 recursion limit, so you can't continue resetting indefinitely, and it also counts as another worker invocation

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:27 PM

Got it so I could theoretically get up to 32,000 invocations in one go

KDX_MW | Kaddex.xyz / Ecko.xyz•12/17/23, 7:27 PM

Presuming I recursively called at the end

Constantin•12/18/23, 12:49 AM

is there a problem with workers deployment/dns

Constantin•12/18/23, 12:52 AM

the wroker is beeing deployed via wrangler having this config

{
    "compatibility_date": "2023-12-06",
    "main": "src/index.ts",
    "account_id": "1cca8d8eb0cc2de3f76e07b7164f222e",
    "placement": {
        "mode": "smart"
    },
    "env": {
        "main": {
            "name": "worker-app-proxy",
            "route": {
                "zone_id": "0ae46da72468555b0841b3615fc6f14e",
                "pattern": "*.app.nefele.ai/*"
            }
        }
    }
}

{
    "compatibility_date": "2023-12-06",
    "main": "src/index.ts",
    "account_id": "1cca8d8eb0cc2de3f76e07b7164f222e",
    "placement": {
        "mode": "smart"
    },
    "env": {
        "main": {
            "name": "worker-app-proxy",
            "route": {
                "zone_id": "0ae46da72468555b0841b3615fc6f14e",
                "pattern": "*.app.nefele.ai/*"
            }
        }
    }
}

all good, however there's no Worker DNS record, and the url is not reacheable

kian•12/18/23, 1:03 AM

Did you create a DNS record for that route?

kian•12/18/23, 1:03 AM

Workers won’t do that for you unless you’re using custom domains.

Constantin•12/18/23, 1:13 AM

yep, i did ..

CConstantin yep, i did ..

Chaika•12/18/23, 1:22 AM

Are you saying you have a wildcard on nefele.ai pointing to app.nefele.ai, and thus you expect it to work for *.app.nefele.ai*.app.nefele.ai?
If so, wildcards are a bit fun with DNS. It'll stop if it hits anything, and stop on any deeper labels too. Meaning, your wildcard stops recursing because you have a record created for app.nefele.ai.
You'd have to create another wildcard record for *.app.nefele.ai*.app.nefele.ai
That's also too deep for your free Universal certificate which only covers first level subdomains, you'd need Advanced Cert Manager (ACM)

Advany•12/18/23, 10:34 AM

will there ever be a new version of the cloudflare apps? I really though it was a powerfull concept!

piffie•12/18/23, 11:26 PM

Hey guys, we are getting updates from around 4k pricepoints, typically around 30-40 per second.
I first thought about running a worker for this, then maybe to query.
But then i reread the durable object API - and it seems i would best actually handle it inside a durable object to hold the values in memory, and give it from there access to others. Whats the best architecture for that?

Unsmart•12/18/23, 11:27 PM

Youre going to have to explain a bit more about what you are looking for if you want anything useful. With only that limited info no one can suggest anything really.

its so over•12/19/23, 7:41 AM

is it possible to make udp connections from a worker

Iits so over is it possible to make udp connections from a worker

Hello, I’m Allie!•12/19/23, 7:42 AM

its so over•12/19/23, 7:42 AM

njhjh•12/19/23, 11:32 AM

Hi all

I am wondering about Sentry Tails Worker pricing. Will attaching Sentry mean that I will get all the requests and pricing doubled? Because Tail workers are priced the same as workers and are automatically invoked after the invocation of a producer Worker

jacob•12/19/23, 1:43 PM

While it's possible to get a user's ip country from a worker, is it possible to get the context of where the worker itself is being run? An api I'm quering has different base urls depending on location, ideally I'd wanna call the region closest to the worker when it's being run

Jjacob While it's possible to get a user's ip country from a worker, is it possible to ...

Isaac McFadyen•12/19/23, 1:46 PM

The only way is to hit a /cdn-cgi/trace/cdn-cgi/trace endpoint and look for the colocolo field. So for example, https://cloudflare.com/cdn-cgi/trace shows EWR and EWR = New York airport, so Worker is being run in New York.

Isaac McFadyen•12/19/23, 1:47 PM

But also it's probably not necessary. Workers run in every datacenter on the Cloudflare network (350+) which means it will almost always be run close to your user.

Isaac McFadyen•12/19/23, 1:47 PM

So if you use the user's country/region it's almost guaranteed to be near the Worker.

IIsaac McFadyen So if you use the user's country/region it's almost guaranteed to be near the Wo...

jacob•12/19/23, 1:48 PM

Thanks

Chaika•12/19/23, 2:11 PM

You could also just get request.cf.colo if you just wanted the datacenter in a Worker, doesn't work in DOs though

CChaika You could also just get request.cf.colo if you just wanted the datacenter in a W...

Isaac McFadyen•12/19/23, 2:29 PM

Oh yeah true lul

kian•12/19/23, 3:36 PM

Should be the original

kian•12/19/23, 3:36 PM

pages-discussions

sanetram•12/19/23, 4:42 PM

Hey, hopefully this is allowed, but would anyone be interested in me releasing this codegen tool for the fastest possible 0-dep routers?

sanetram•12/19/23, 4:44 PM

aim was to generate switch cases with string literals to use v8 parse-time optimisations. it's also meant to be readable so i'm not custom rolling any "few"-case minifications into if statements etc and move that to the bundler

juri•12/19/23, 6:51 PM

Does anyone have a reliable way to track timings inside of a worker? I am aware that Date.nowDate.now and performance.nowperformance.now are only ever updated when I/O occurs, but how do you guys go about debugging internal latency issues?

just to be accurate, KV reads are $0.50/M, and writes/lists/deletes are $5/M

Similar Threads

Similar Threads

Similar Threads