Guys, why the hell is metadata a required field? What should I pass there? Because I know what "Body

L

LukeZOP•7/14/25, 6:06 PM

LLukeZ Guys, why the hell is metadata a required field? What should I pass there? Becau...

I

Isaac McFadyen•7/14/25, 6:29 PM

If you don't have any metadata, set to the empty object

{}

{}

I

Isaac McFadyen•7/14/25, 6:29 PM

It's basically extra data to be attached to the KV entry that you can see in a

list

list

list

list call if you need to

L

LukeZOP•7/14/25, 6:29 PM

I have to send an empty string, but thanks!

IIsaac McFadyen It's basically extra data to be attached to the KV entry that you can see in a `...

L

LukeZOP•7/14/25, 6:30 PM

aaah, interesting

LLukeZ Guys, why the hell is metadata a required field? What should I pass there? Becau...

T

thomasgauvin•7/15/25, 1:05 PM

It's needed in the REST API (can be empty object like Isaac mentioned) but is there a reason you are using the REST API instead of the binding?

L

LukeZOP•7/15/25, 1:19 PM

because it's not a worker and I don't have my app deployed on workers because it is too much work for this simple, single use case

T

Tony Leung•7/22/25, 6:25 PM

hi guys, will Cloudflare eventually support bulk delete or bulk put in cloudflare KV using in Workers? CF workers has a limitation of 6 concurrent connection at the same time and it's really painful to bulk put like 20 - 30 items using the SDK in workers.

TTony Leung hi guys, will Cloudflare eventually support bulk delete or bulk put in cloudflar...

H

Hard@Work•7/22/25, 6:45 PM

Maybe put your question in #community-q-and-a ?

HHard@Work Maybe put your question in #community-q-and-a ?

T

Tony Leung•7/22/25, 6:45 PM

thanks, chaika has already answered it

sorry for posting it off-topic

H

Hth4nh•7/24/25, 7:05 PM

Hello everyone,

I'm encountering an issue with the Cloudflare Worker KV where I get an empty array when trying to retrieve multiple keys at once.

When I run the following code, it returns an empty array:

const values = await options.env.TEST.get(["2930", "2929"]);
console.log([...values.entries()]); // []

// Expected: [["2930", "foo"],["2929","bar"]]

const values = await options.env.TEST.get(["2930", "2929"]);
console.log([...values.entries()]); // []

// Expected: [["2930", "foo"],["2929","bar"]]

const values = await options.env.TEST.get(["2930", "2929"]);
console.log([...values.entries()]); // []

// Expected: [["2930", "foo"],["2929","bar"]]

const values = await options.env.TEST.get(["2930", "2929"]);
console.log([...values.entries()]); // []

// Expected: [["2930", "foo"],["2929","bar"]]

However, this alternative approach works as expected:

const values = await Promise.all(["2930", "2929"].map((key) => options.env.TEST.get(key)));
console.log(values); // ["foo", "bar"]

const values = await Promise.all(["2930", "2929"].map((key) => options.env.TEST.get(key)));
console.log(values); // ["foo", "bar"]

const values = await Promise.all(["2930", "2929"].map((key) => options.env.TEST.get(key)));
console.log(values); // ["foo", "bar"]

const values = await Promise.all(["2930", "2929"].map((key) => options.env.TEST.get(key)));
console.log(values); // ["foo", "bar"]

According to the documentation, passing an array of keys to the

get()

get()

get()

get() method should return a

Map

Map

Map

Map of the key-value pairs.

Here is my

wrangler.jsonc

wrangler.jsonc

wrangler.jsonc

wrangler.jsonc file for reference:

{
    "$schema": "node_modules/wrangler/config-schema.json",
    "name": "***",
    "main": "./src/index.ts",
    "compatibility_date": "2025-06-20",
    "observability": {
        "enabled": true
    },
    "assets": {
        "directory": "./public/",
        "binding": "ASSETS"
    },
    "kv_namespaces": [
        {
            "binding": "TEST",
            "id": "***",
            "preview_id": "***"
        }
    ]
}

{
    "$schema": "node_modules/wrangler/config-schema.json",
    "name": "***",
    "main": "./src/index.ts",
    "compatibility_date": "2025-06-20",
    "observability": {
        "enabled": true
    },
    "assets": {
        "directory": "./public/",
        "binding": "ASSETS"
    },
    "kv_namespaces": [
        {
            "binding": "TEST",
            "id": "***",
            "preview_id": "***"
        }
    ]
}

{
    "$schema": "node_modules/wrangler/config-schema.json",
    "name": "***",
    "main": "./src/index.ts",
    "compatibility_date": "2025-06-20",
    "observability": {
        "enabled": true
    },
    "assets": {
        "directory": "./public/",
        "binding": "ASSETS"
    },
    "kv_namespaces": [
        {
            "binding": "TEST",
            "id": "***",
            "preview_id": "***"
        }
    ]
}

{
    "$schema": "node_modules/wrangler/config-schema.json",
    "name": "***",
    "main": "./src/index.ts",
    "compatibility_date": "2025-06-20",
    "observability": {
        "enabled": true
    },
    "assets": {
        "directory": "./public/",
        "binding": "ASSETS"
    },
    "kv_namespaces": [
        {
            "binding": "TEST",
            "id": "***",
            "preview_id": "***"
        }
    ]
}

I've also made sure that the

preview_id

preview_id

preview_id

preview_id is correctly configured for use with

wrangler dev

wrangler dev

wrangler dev

wrangler dev.

Has anyone else experienced this issue or has any idea what might be going wrong?

Thanks

M

mr.niko.la•8/1/25, 2:28 AM

Is kv able to temp hold 25mb file ?

For <60s

M

mr.niko.la•8/1/25, 2:28 AM

I need failover for r2 upload and prefer to use cf infra for latency before trying another s3

Mmr.niko.la Is kv able to temp hold 25mb file ? For <60s

I

Isaac McFadyen•8/1/25, 4:25 AM

25mb is the limit per KV value. It can hold any value up to 25mb indefinitely, yes.

P

Pingu•8/5/25, 5:07 PM

Do KV key expirations count as deletes for billing purposes?

C

cosbgn•8/7/25, 8:45 AM

Is KV down?

Ccosbgn Is KV down?

I

Isaac McFadyen•8/7/25, 1:38 PM

Are you still seeing these issues?

Original message was deleted

H

Hard@Work•8/7/25, 6:34 PM

?pings

S

SuperHelpflare•8/7/25, 6:34 PM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

IIsaac McFadyen Are you still seeing these issues?

C

cosbgn•8/7/25, 6:46 PM

Last was 21h ago

Y

yevgen•8/12/25, 9:25 AM

In case you missed it last week: https://blog.cloudflare.com/rearchitecting-workers-kv-for-redundancy/

The Cloudflare Blog

Redesigning Workers KV for increased availability and faster perfor...

Workers KV is Cloudflare's global key-value store, serving millions of requests per second across hundreds of billions of stored objects. The service powers critical infrastructure for dozens of Cloudflare products—from Access authentication to Pages static assets—making its availability essential to our platform's reliability.

After the in...

A

alison•8/18/25, 8:50 PM

There's a relatively well-documented issue with writing to KV where

metadata

metadata

is supposedly optional, but required in the type definition within the

cloudflare-typescript

cloudflare-typescript

cloudflare-typescript

cloudflare-typescript library.

Easy, right? Just pass in a dummy empty

metadata

metadata

object on your writes to KV! WRONG!!

If you just do that, your value gets stored in KV as an object like

{ value: XYZ, metadata: {}}

{ value: XYZ, metadata: {}}

{ value: XYZ, metadata: {}}

{ value: XYZ, metadata: {}}. Not super intuitive, I'd rather just store XYZ. So then I learned:

When passing in

metadata

metadata

, you actually need to include the

Content-Type: multipart/form-metadata

Content-Type: multipart/form-metadata

Content-Type: multipart/form-metadata

Content-Type: multipart/form-metadata header (image 1, link). The library has some handling for automatically including this header, but it doesn't catch these cases where we're forced to pass in empty metadata

Logical next move is to just pass in the header. Ok cool, easy right? WRONG!

Now I get an error back:

{"code":10047,"message":"can not parse value and metadata from multipart request body: 'could not parse multipart request: 'no multipart boundary param in Content-Type''"}

{"code":10047,"message":"can not parse value and metadata from multipart request body: 'could not parse multipart request: 'no multipart boundary param in Content-Type''"}

{"code":10047,"message":"can not parse value and metadata from multipart request body: 'could not parse multipart request: 'no multipart boundary param in Content-Type''"}

{"code":10047,"message":"can not parse value and metadata from multipart request body: 'could not parse multipart request: 'no multipart boundary param in Content-Type''"}

Fair, but I feel I am entering hacky code land and am not exactly sure how to set and use a boundary param within the context of the API. Is it as easy as just passing something like

Content-Type: multipart/form-data; boundary=asfghjkl

Content-Type: multipart/form-data; boundary=asfghjkl

Content-Type: multipart/form-data; boundary=asfghjkl

Content-Type: multipart/form-data; boundary=asfghjkl?

Has anyone run into this issue? If so, how did you fix it? Did you just take the L and take

.value

.value

.value

.value on your GET response from KV, potentially running into issues when this finally gets fixed? Or did you manually pass in a fancy header? Or something else I am forgetting?

Here are some previous issues where people encountered the same thing: https://github.com/cloudflare/cloudflare-go/issues/3911, https://github.com/cloudflare/cloudflare-python/issues/2519

S

shinshin•8/25/25, 9:52 AM

When calling get using a key obtained from the list, null may be returned even if the key exists. Is this behavior caused by KV consistency?

const keys = await env.NAMESPACE.list();

Promise.all(keys.map(async (key) => {
  const v = env.NAMESPACE.get(key.name);
  console.log(v);
}))

const keys = await env.NAMESPACE.list();

Promise.all(keys.map(async (key) => {
  const v = env.NAMESPACE.get(key.name);
  console.log(v);
}))

const keys = await env.NAMESPACE.list();

Promise.all(keys.map(async (key) => {
  const v = env.NAMESPACE.get(key.name);
  console.log(v);
}))

const keys = await env.NAMESPACE.list();

Promise.all(keys.map(async (key) => {
  const v = env.NAMESPACE.get(key.name);
  console.log(v);
}))

Sshinshin When calling get using a key obtained from the list, null may be returned even i...

I

Isaac McFadyen•8/25/25, 3:43 PM

Depends, if the key was recently written then yes.

I

Isaac McFadyen•8/25/25, 3:43 PM

If the key has been present for a while you should see it fine though.

J

João Castro•8/25/25, 11:26 PM

I don't understand much about how KV is built, but would having something like https://blog.cloudflare.com/instant-purge/ possible or would it break the fundamentals of how it works? Sometimes the 60s delay for revalidating keys can be pretty annoying

The Cloudflare Blog

Instant Purge: invalidating cached content in under 150ms

Today we’re excited to share that we’ve built the fastest cache purge in the industry. We now offer a global purge latency for purge by tags, hostnames, and prefixes of less than 150ms on average (P50), representing a 90% improvement since May 2022.

IIsaac McFadyen 25mb is the limit per KV value. It can hold any value up to 25mb indefinitely, y...

M

mr.niko.la•8/27/25, 1:12 PM

Basically in a hyper rare case r2 upload fails due some reason can kv be used as failover ? Is that even worth doing or is cf automatically doing that? Or is it better to use something like backblaze ?

S

ShaneCodes•8/28/25, 1:27 PM

We use a ton of KV at Zephyr and I'm always poking around local KVs and was having a really hard time find the resources with the way they're stored. The SQLite table which has the actual keys mapped to the blob id. So I made a vscode plugin to connect the two and let me search them. I don't know if anyone else would find it useful but wanted to share

https://marketplace.visualstudio.com/items?itemName=swalkerdev.cloudflare-kv-explorer

Cloudflare KV Explorer - Visual Studio Marketplace

Extension for Visual Studio Code - Browse and inspect Cloudflare Workers KV stores during local development with Wrangler

S

ShaneCodes•8/28/25, 1:28 PM

I don't know if the dev tools are open source but I would really be down to add this functionality to the dev tools if I can

L

LukeZOP•9/1/25, 5:17 AM

How much workload can a KV handle? How to make writes more efficient?

I'm planning on making a website where you can basically press just one button, but I want to keep track of how many times people pressed the button.

However, I know there are certain limits to the KV (reads & writes) so basically just getting the value and incrementing it would be too inefficient since it would be two requests to the KV every time the button is clicked.
So what would you guys suggest to make this more efficient?

LLukeZ ### How much workload can a KV handle? How to make writes more efficient? I'm p...

I

Isaac McFadyen•9/1/25, 5:18 AM

just getting the value and incrementing it would be too inefficient since it would be two requests to the KV every time the button is clicked.

Not only that, KV is eventually consistent. If you do a

get()

get()

get()

get() in one datacenter and then increment and a

put()

put()

put()

put(), another datacenter might not see it for up to 60 seconds so that datacenter might also do a

get()

get()

get()

get() and a

put()

put()

put()

put() (using the stale value) and then your counter is no longer consistent since a write would be lost.

I

Isaac McFadyen•9/1/25, 5:19 AM

I'd recommend another storage solution like Durable Objects or D1 for this use-case. KV is great for read-heavy workloads but not a good fit for workloads that need strong consistency like this.

IIsaac McFadyen > just getting the value and incrementing it would be too inefficient since it w...

L

LukeZOP•9/1/25, 5:20 AM

Oh! Interesting to know

IIsaac McFadyen I'd recommend another storage solution like Durable Objects or D1 for this use-c...

L

LukeZOP•9/1/25, 5:21 AM

Aight

I will take a look at this, thanks!

E

Eumatio•9/1/25, 2:54 PM

Hii! I wanted to know what is the reason for this limit, or if is just a bug?

const state = ulid(Date.now());

const stateData: StateData = {
    returnTo: safeReturn,
    codeVerifier: verifier,
    createdAt: Date.now(),
    expiresAt: Date.now() + this.STATE_TTL_MS
};

await this.stateDatabase.set(state, JSON.stringify(stateData), {
    expirationTtl: this.STATE_TTL_MS
});

const state = ulid(Date.now());

const stateData: StateData = {
    returnTo: safeReturn,
    codeVerifier: verifier,
    createdAt: Date.now(),
    expiresAt: Date.now() + this.STATE_TTL_MS
};

await this.stateDatabase.set(state, JSON.stringify(stateData), {
    expirationTtl: this.STATE_TTL_MS
});

const state = ulid(Date.now());

const stateData: StateData = {
    returnTo: safeReturn,
    codeVerifier: verifier,
    createdAt: Date.now(),
    expiresAt: Date.now() + this.STATE_TTL_MS
};

await this.stateDatabase.set(state, JSON.stringify(stateData), {
    expirationTtl: this.STATE_TTL_MS
});

const state = ulid(Date.now());

const stateData: StateData = {
    returnTo: safeReturn,
    codeVerifier: verifier,
    createdAt: Date.now(),
    expiresAt: Date.now() + this.STATE_TTL_MS
};

await this.stateDatabase.set(state, JSON.stringify(stateData), {
    expirationTtl: this.STATE_TTL_MS
});

W

Walshy•9/1/25, 5:13 PM

that looks like your STALE_TTL_MS is wrong

W

Walshy•9/1/25, 5:13 PM

it shouldn't be over/under 2 billion milliseconds

Aalison There's a relatively well-documented issue with writing to KV where `metadata` i...

T

thomasgauvin•9/3/25, 2:19 PM

Hey @alison & folks, we've recently released the v5.0.0 version of the

cloudflare

cloudflare

cloudflare

cloudflare typescript SDK, which resolves the metadata bug reported above with workers kv updates and reads. You are now able to write, update & read the metadata from the typescript SDK! Go try it out

https://www.npmjs.com/package/cloudflare/v/5.0.0

npm

cloudflare

The official TypeScript library for the Cloudflare API. Latest version: 5.0.0, last published: 18 hours ago. Start using cloudflare in your project by running

npm i cloudflare

npm i cloudflare

npm i cloudflare

npm i cloudflare. There are 124 other projects in the npm registry using cloudflare.

S

Sb•9/9/25, 1:48 PM

Hello,

Does reading and writing from the meta of the kv key-value count in the billing ?

SSb Hello, Does reading and writing from the meta of the kv key-value count in the ...

I

Isaac McFadyen•9/9/25, 1:49 PM

In order to access KV metadata you need to either make a

list()

list()

list()

list() call (which is billed) or a

get()

get()

get()

get()/

getWithMetadata()

getWithMetadata()

getWithMetadata()

getWithMetadata() call (which is also billed). So yes.

S

Sb•9/9/25, 1:56 PM

Okay,

I want to make a counter which counts the no. of requests hitting my worker. Is there a better option for this value fetch ??

I was thinking of kv earlier but it is billable and expensive for a simple counter property as it requires write action in every request.

Is there a better way for this value to be accessible in the worker itself?

SSb Okay, I want to make a counter which counts the no. of requests hitting my work...

I

Isaac McFadyen•9/9/25, 1:57 PM

KV is not a good fit for that anyway because it's eventually consistent. See my message here: https://canary.discord.com/channels/595317990191398933/874727019479437372/1411943337698529381

I'd recommend Durable Objects or D1, both of which will likely be cheaper.

S

Sb•9/9/25, 2:03 PM

I saw firebase real-time as one of the good fit in this case, would you recommend using realtime in place of kv in this situation?

I

Isaac McFadyen•9/9/25, 2:11 PM

Firebase pricing is quite high, so depending on Worker traffic you'd quickly hit the limit of a free tier and be charged.

I

Isaac McFadyen•9/9/25, 2:12 PM

It's also not very fast, meaning you'd need to use

waitUntil()

waitUntil()

waitUntil()

waitUntil() to avoid delaying the request.

SSb Okay, I want to make a counter which counts the no. of requests hitting my work...

1

1534980751438970153470891543•9/9/25, 2:22 PM

#analytics-engine also good for this & non blocking

SSb Okay, I want to make a counter which counts the no. of requests hitting my work...

T

thomasgauvin•9/9/25, 3:16 PM

why do you need to count the amount of requests hittin gyour worker? why is worker analytics insufficient?

Tthomasgauvin why do you need to count the amount of requests hittin gyour worker? why is work...

S

Sb•9/9/25, 3:29 PM

I want to set force stop on the worker if it faces massive amount of requests to keep it within the client's budget.

SSb I want to set force stop on the worker if it faces massive amount of requests to...

I

Isaac McFadyen•9/9/25, 3:37 PM

If I'm understanding what you're asking, the worker will always be invoked. In order to 'stop' the Worker you'd need to unbind it from the actual domain/zone, because once a request reaches far enough for you to check if you're over the limit it would be billed.

IIsaac McFadyen If I'm understanding what you're asking, the worker will always be invoked. In o...

S

Sb•9/9/25, 5:13 PM

So, what i have configured is.

Instead of binding the worker to the zone via zone property, I added a route to the domain in the worker configuration.

And when I hit (may be a X number) of requests, I would like to stop serving on that route or remove those routes.

What I remember from the free tier is, it stops automatically after 100k and I want to set similar kinda concept in the paid tier also.

I am using pages to host static files mainly (as any incorrect automatically triggers 404) and using worker for (some static assets), dynamic pages, and api endpoints.

When the client set limit exceeds, I don't want the workers to be invoked on those routes to save clients cost. (Stop api.xyz.com)

Guys, why the hell is metadata a required field? What should I pass there? Because I know what "Body

How much workload can a KV handle? How to make writes more efficient?

Similar Threads

Similar Threads

How much workload can a KV handle? How to make writes more efficient?

Similar Threads