KV caching is the Cache API under the hood, iirc

HHello, I’m Allie!KV caching is the Cache API under the hood, iirc

Rich•5/9/24, 12:46 PM

I did wonder that but the docs suggest it takes advantage of tiered caching which the Cache API doesn't

Data must be read from the nearest regional tier, followed by a central tier, degrading finally to the central store for a truly cold global read

Rich•5/9/24, 12:50 PM

The way the docs describe it, I was hoping the KV Store would act similarly to Tiered Caching and Cache Reserve -- neither of which you get with the Cache API

RRich I did wonder that but the docs suggest it takes advantage of tiered caching whic...

Chaika•5/9/24, 1:18 PM

the docs are wrong, they were updated to talk about KV 2.0 (or whatever you want to call it), but that was reverted and pulled back

Chaika•5/9/24, 1:21 PM

they pulled the blog post too, this is what KV 2.0 was supposed to be: https://web.archive.org/web/20230701095201/https://blog.cloudflare.com/faster-workers-kv-architecture/

Chaika•5/9/24, 1:22 PM

we raised an issue about getting those docs reverted a while ago but I guess nothing came of it

Chaika•5/9/24, 1:24 PM

(but regardless, the way the current kv works is just per location cache and then one of the central stores. KV's cache is also slower by ~20ms or so then normal cache api I've noticed, not sure what it is exactly under the hood but it behaves pretty much the same as the cache api besides the slowness)
here's the older doc which reflects the current released version of kv: https://web.archive.org/web/20230520041620/https://developers.cloudflare.com/workers/learning/how-kv-works/

kian•5/9/24, 1:30 PM

I imagine that the KV cache is implemented in KV's Worker, so you still have to go talk to that

kian•5/9/24, 1:30 PM

idk why else it'd be slower :blobhaj_thonk:

Kkian I imagine that the KV cache is implemented in KV's Worker, so you still have to ...

Chaika•5/9/24, 1:30 PM

yea but that wouldn't be 20-30ms slower just to talk to a worker one would think

kian•5/9/24, 1:34 PM

If Date.now() for IO is to be believed, my Worker fetching a Hello World Worker takes about 15-25ms

kian•5/9/24, 1:34 PM

It wouldn't surprise me if it is that

Rich•5/9/24, 1:40 PM

Yikes, that's really disappointing. Thanks for the info.

I'm coming to the realisation that CF is only really beneficial for very high traffic sites -- I wonder if all CDNs are as quick to evict cache entries.

shrut•5/13/24, 2:00 PM

hey community, this is my issue.
Everytime i deploy my application bun run deploybun run deploy then all the enviroment variables in the wrangler.toml file are directly goes through very easy, can i do the same with the kv_namespaces because it is a tedious job to set the kv_namespace in the settings everytime from the cloudflare dashboard when we have many kv_namespaces and the application is in production as well as under constant development so everytime i deploy it its hard to write all the kv_namespace variables, its get messy especially when many people are working on it at the same time i hope you get the point
thank you

Sshrut hey community, this is my issue. Everytime i deploy my application `bun run dep...

Hello, I’m Allie!OP•5/13/24, 2:13 PM

https://developers.cloudflare.com/workers/wrangler/configuration#kv-namespaces

Cloudflare Docs

Configuration - Wrangler · Cloudflare Workers docs

Use a

wrangler.toml

wrangler.toml

configuration file to customize the development and deployment setup for your Worker project and other Developer Platform …

Configuration - Wrangler · Cloudflare Workers docs

Sasha 🐑💨 Koss•5/18/24, 2:13 AM

Hey! I have a question. I want to use multi-tenant architecture and create separate D1 databases for each organization, and its limits (50K per account) seem to encourage this approach.

However, I can see that KV allows only 100 namespaces on paid accounts. Is creating a namespace for each tenant anti-pattern, or is it just an artificial limitation, e.g., for Cloudflare to learn about their customers' needs?

FTXflare•5/18/24, 10:18 PM

Is accessing cache on the front to KV count as read operation?

Cyb3r-Jak3•5/19/24, 12:59 PM

The cache API is free

CCyb3r-Jak3 The cache API is free

Hello, I’m Allie!OP•5/19/24, 1:53 PM

Though if you mean a cached value from KV directly, then that is billed

HHello, I’m Allie!Though if you mean a cached value from KV **directly**, then that is billed

FTXflare•5/20/24, 12:47 PM

yes, I was asking about that, bc as I understand it worker is on the front to cache

SSasha 🐑💨 Koss Hey! I have a question. I want to use multi-tenant architecture and create separ...

thomasgauvin•5/20/24, 1:50 PM

Are you asking specifically about KV? What's you use case for multi-tenant architecture using KV? Feel free to DM me if you can't share the details. Building multi-tenant applications with KV is currently done by most the traditional way, with a single instance

arbfay•5/20/24, 6:20 PM

Hey I hope I don't sound too annoying talking about performance of KV.
It's just that I've been using KV for a lot of "indexing" tasks (to map resources to DOs, databases, etc), and for processes (oauth, etc). Often I'd have to read 2 or 3 keys then request a DO, then maybe do a write in KV, etc. But it's... slow. That can take several seconds (in Europe), sometimes > 10 secs, not good to provide a snappy experience at all. And I just discovered KV might be the slowest too... (according to https://deno.com/blog/comparing-deno-kv, but I don't trust benchmarks anyway). It feels like using DO automatically means losing any latency gain you'd think to get from deploying at the edge.

My ask is, are we going to hear about some performance improvements at some point?

Unsmart•5/20/24, 6:27 PM

That specific benchmark has already been debunked of any value. For DenoKV they do a simple read op, for every other service they do a list followed by reads thus the higher latency.

But yeah KV is not the fastest for when you need to hit origin for cold reads or for writes.

thomasgauvin•5/21/24, 2:28 PM

@arbfay just reached out to you, usually KV is highly performant for read heavy workloads but I want to better understand your use case

Jitendra•5/23/24, 10:57 AM

Hello everyone.

I need your opinion on using KV. I am using KV for cache as well as for rate limiting the email sending attempt.

Would you prefer it for rate limiting?? Before this I was using Redis. But I have decided to go all in Cloudflare

Also I am facing some issue

Like I want to limit total email count in a day to 10.
Focusing on it like I couldn't find any update function so how to update.
If I couldn't update that means it would change the TTL right??

JJitendra Hello everyone. I need your opinion on using KV. I am using KV for cache as wel...

Hello, I’m Allie!OP•5/23/24, 11:38 AM

KV is not a good idea for ratelimiting, since it is eventually consistent. I would go with a Durable Object instead

HHello, I’m Allie!KV is not a good idea for ratelimiting, since it is eventually consistent. I wou...

Jitendra•5/23/24, 11:39 AM

Thanks.

Walshy•5/23/24, 11:53 AM

Even better, the rate limit binding

WWalshy Even better, the rate limit binding 🙂

Hello, I’m Allie!OP•5/23/24, 12:04 PM

Note that this only applies to the datacenter/colo the requests hit, so if an account can trigger sends from multiple locations, they will not be applied to the same ratelimit.

[

[BR]Edson•5/24/24, 12:53 PM

anyone can help me underestand one point? im storing the url path as key of kv todo a rudimentary cache layer for some responses. but when i search on the GUI, none of the stored data shows on, it works fine on code but i cant check on the GUI. Theres any limitation for search keys started with / or having any special caracteres?

Kurzon•5/25/24, 5:37 AM

KV Namespace itself works but the key/value are null?

Kurzon•5/25/24, 5:37 AM

I'm running this in

wrangler dev

wrangler dev

Kurzon•5/25/24, 5:37 AM

This prints

null

null

Kurzon•5/25/24, 5:41 AM

Never mind, local development does not allow access, have to create a separate namespace in CLI.

Kurzon•5/25/24, 5:42 AM

But I never saw this in any of the Cloudflare Workers docs, it should be added into at least one of them for KV, I use the docs a lot. It is probably in one of them but I found my answer for this through stack overflow, not through the docs.

Puliczek•5/26/24, 2:58 PM

is it a good practice to store JSON as a value in KV? for example key:"user_123", value:"{"credits":15,email:"az@gmail.com"} ?

PPuliczek is it a good practice to store JSON as a value in KV? for example key:"user_123"...

Hello, I’m Allie!OP•5/26/24, 3:01 PM

Yeah, that's fine. In fact, when reading a value from KV, you can await KV.get("key, "json")await KV.get("key, "json") to have KV automagically parse it for you

HHello, I’m Allie!Yeah, that's fine. In fact, when reading a value from KV, you can `await KV.get(...

Puliczek•5/26/24, 3:02 PM

cool, thanks

cosbgn•5/27/24, 11:05 AM

KV is giving me quite a bit of erorr 500500 - Is there anything I can do? Is this normal?

cosbgn•5/27/24, 11:05 AM

for the past few weeks I get once/twice a day almost

raj.js•5/31/24, 8:19 PM

need a suggestion:
Need to have a counter but using kv can be an issue as its say only 1 write per second to same key. is there anyway around it? we need high frequent concurrent atleast 1000 writes per second (basically one write per read). is there anyway to do it using KV or have to go with redis like cache

Mrinank•5/31/24, 8:49 PM

Why kv doesn't work in typescript code. I am not getting error in js files, having problem with env in ts

MMrinank Why kv doesn't work in typescript code. I am not getting error in js files, havi...

Mrinank•5/31/24, 8:55 PM

Cannot read properties of undefined (reading 'put')

MMrinank Why kv doesn't work in typescript code. I am not getting error in js files, havi...

Matt Silverlock•6/2/24, 9:39 PM

Can you share the code, including how you have defined your bindings?

MMatt Silverlock Can you share the code, including how you have defined your bindings?

Mrinank•6/3/24, 7:29 PM

Its working now

Mrinank•6/3/24, 7:30 PM

sharing the code for others

name = "newcron"
main = "src/index.ts"
compatibility_date = "2024-05-29"


[triggers]
crons = ["* */12 * * *"]

[[kv_namespaces]]
binding = "KV"
id = "b6783170bcde4c6f9d9e6152dc8d1027"

[ai]
binding = "AI"

name = "newcron"
main = "src/index.ts"
compatibility_date = "2024-05-29"


[triggers]
crons = ["* */12 * * *"]

[[kv_namespaces]]
binding = "KV"
id = "b6783170bcde4c6f9d9e6152dc8d1027"

[ai]
binding = "AI"

AKStriker98•6/4/24, 10:06 PM

Is there a way to deal with the graveyard that is created from unread items being deleted by the TTL? Had about 5 GB of unread data expire and now we cannot get to the valid data with list because it’s stuck churning through the graveyard. We don’t want to delete and recreate the kv store since it’s tied to multiple apis inputting data to them and being read by workers

KV caching is the Cache API under the hood, iirc

Similar Threads

Similar Threads

Similar Threads