What's the reason to have a 1000 api request limit per worker ? -- Context: I built a global rate

SShivek What's the reason to have a 1000 api request limit per worker ? -- Context: I ...

P

Peps•10/7/25, 1:26 PM

1000 subrequests can be made per invocation, so basically each time the worker is hit. It's not in total for the entire lifespan of the worker

P

Peps•10/7/25, 1:26 PM

unless you need to send 1000 requests to third parties for every single request that hits your worker, it won't be a problem

P

Peps•10/7/25, 1:28 PM

https://developers.cloudflare.com/workers/platform/limits/#how-many-subrequests-can-i-make

Mm0c Aloha, is there an easy way to find all logs from a workflow instance? I think i...

A

avenceslau•10/7/25, 2:53 PM

Not currently no, sorry about that. Its on our roadmap we will get to it :soontm:

A

avenceslau•10/7/25, 3:00 PM

I find that the best way to debug a workflow is to run is to add additional steps and use their results messages as logs (this is my what I personally what I do when I do need to have more immediate logs). Logs do appear in workers observability but only when the workflow instance stops running which is not ideal.

Mm0c Aloha, is there an easy way to find all logs from a workflow instance? I think i...

A

avenceslau•10/7/25, 3:08 PM

This is not ideal but will get you live-ish logs for an instance (I don't recommend having this in production but while you test, it can be helpful)

await step.do("my-debug-step", async () => {
  return 'this is how I log, and it will show up as the result of this step'
})

const myResult = await step.do() // this step returns something that we want to log
await step.do("my-debug-step", async () => {
  if(myResult) {
    return `You can also log results ${JSON.stringify(myResult)}`
  }
  else {
    return `Result is not defined`
  }  
})

await step.do("my-debug-step", async () => {
  return 'this is how I log, and it will show up as the result of this step'
})

const myResult = await step.do() // this step returns something that we want to log
await step.do("my-debug-step", async () => {
  if(myResult) {
    return `You can also log results ${JSON.stringify(myResult)}`
  }
  else {
    return `Result is not defined`
  }  
})

await step.do("my-debug-step", async () => {
  return 'this is how I log, and it will show up as the result of this step'
})

const myResult = await step.do() // this step returns something that we want to log
await step.do("my-debug-step", async () => {
  if(myResult) {
    return `You can also log results ${JSON.stringify(myResult)}`
  }
  else {
    return `Result is not defined`
  }  
})

await step.do("my-debug-step", async () => {
  return 'this is how I log, and it will show up as the result of this step'
})

const myResult = await step.do() // this step returns something that we want to log
await step.do("my-debug-step", async () => {
  if(myResult) {
    return `You can also log results ${JSON.stringify(myResult)}`
  }
  else {
    return `Result is not defined`
  }  
})

B

BlockMinister•10/7/25, 3:39 PM

Hey guys, I'm seeing a crazy number of random "Attempt failed due to internal workflows error" effects in various different Workflow steps. How can i get further info as to whats causing it? I've trawled the logs but can see nothing that would indicate anything abnormal.

M

Matt Silverlock•10/9/25, 8:52 PM

https://developers.cloudflare.com/changelog/2025-10-09-workflows-terraform/

Cloudflare Docs

You can now deploy and manage Workflows in Terraform

You can now deploy and manage Workflows in Terraform, using the Cloudflare Terraform provider v5.11.0.

PPeps 1000 subrequests can be made _per invocation_, so basically each time the worker...

S

ShivekOP•10/10/25, 12:24 PM

1000 total invocations.

I'm running a workflow that has many steps.
I need to make 300-400 api calls.
But these calls need to be rate limited, so I added a global limiter Durable Object
Calls to the DO also get counted in the 1000 limit
And I start hitting the invocation limit exceeded
I feel I've hit the wall on what workers platform was designed to do

SShivek 1000 total invocations. - I'm running a workflow that has many steps. - I need...

V

vijayaraagavan•10/10/25, 2:23 PM

I’ve had the same experience. Because of the subrequest limit, I had to split my task across multiple workflow instances and then manually track all their statuses just to determine when the parent job was finished. A 1000 limit makes sense for Workers, but definitely not for Workflows. Hopefully they raise it soon.

E

Eumatio•10/10/25, 8:27 PM

I can no longer access the Workflow page on the Cloudflare dashboard, it disappeared about an hour ago. I’ve already tried clearing my cache and accessing it from a different computer, but the issue persists. The workflow instances are still being created and finalized normally. The issue persists when I am using other network connections

P

Peps•10/10/25, 8:38 PM

known issue with the dashboard, its being worked on right now

EEumatio I can no longer access the Workflow page on the Cloudflare dashboard, it disappe...

H

Hard@Work•10/10/25, 8:46 PM

https://www.cloudflarestatus.com/incidents/7jv9tr6793pd

Some Dashboard Features are Unavailable

O

Olriko•10/14/25, 1:28 PM

Hello, I have an issue. When I run a workflow, the workflow is always queued..

L

Luktaros•10/14/25, 7:30 PM

Hi, does anyone know why I get this error even when my workflow completes succesfully?

[
{
message: 'run',
level: 'error',
$workers: {
event: { rpcMethod: 'run' },
scriptName: '----------',
outcome: 'exception',
eventType: 'rpc',
diagnosticsChannelEvents: [],
entrypoint: '---------,
scriptVersion: { id: '---------------------------- },
truncated: false,
executionModel: 'stateless',
wallTimeMs: 783,
cpuTimeMs: 10,
requestId: '------------------',
},
$metadata: {
id: '--------------------',
requestId: '-----------------',
trigger: 'default.run',
service: '--------------',
level: 'error',
error: 'run',
message: 'run',
account: '---------------------',
provider: 'cloudflare',
type: 'cf-worker-event',
fingerprint: '-----------------',
origin: 'rpc',
messageTemplate: 'run',
},
},
];

I can't catch any error to justify it

SSeekerdasbatatas This is an unfortunate side effect of how we guarantee durable execution, and we...

L

Luktaros•10/14/25, 7:33 PM

Oh, could this be the reason?

LLuktaros Oh, could this be the reason?

A

avenceslau•10/14/25, 7:52 PM

Yup we are working on it

Aavenceslau Yup we are working on it

L

Luktaros•10/14/25, 8:08 PM

Thank you so much for your answer, it was driving me crazy

, also let me add that I'm having a blast working with workflow

F

fredviken•10/14/25, 10:27 PM

would it be possible to use the waitForEvent to make a some sort of debounce?

O

oscar•10/16/25, 11:33 AM

Hello! Seeing a sharp rise on “internal workflows error” once again

D

DevZ•10/17/25, 3:42 PM

Hi guys, I need assistance to fix this issue: Too many API requests by single worker invocation. I have paid workers account. I am using Workflows alongside Queues to fetch, process, and store data.

D

daniel•10/18/25, 5:53 AM

hey is there a way to do per-user queues when triggering Workflows? Either natively within the Workflows or by using Queues to trigger workflows? I would like to be able to have per-user queues when triggering workflows via API but also when creating workflows from within another workflow

T

Tuna•10/18/25, 2:55 PM

Hey everyone!

Can someone help clarify how workflow memory and CPU limits are applied across steps and nested steps?
I’m trying to understand whether:
Each step gets its own CPU and memory allocation, or
Nested steps share the parent’s resource limits.
I tried looking this up in the docs, but the explanation wasn’t very clear. Any insights would be appreciated!

T

TrueBrain•10/19/25, 7:57 PM

Hello!

The Worker attached to a Workflow shows a lot more requests than the amount of invocations to the Workflow. Is this normal? Is this hibernation at work, or some other mechanism?
My workflow has a lot of

sleep("10 minutes")

sleep("10 minutes")

sleep("10 minutes")

sleep("10 minutes"), so I would assume it hibernates a lot

The docs state the pricing of a Workflow is per invocation; so can I ignore the amount of requests the Worker is showing me? Or are both charged? Or how does this work?

Any insights are appreciated

PS: possibly related, but there are also a lot of "cancelled" requests, the metrics tell me. Can't find any details on what "cancelled" means in this context.

N

Need_Not•10/21/25, 11:51 PM

It would be cool if workflows could emit data at each step

K

King Kazu•10/22/25, 10:29 PM

workflows would be great if there was local observability like how inngest has it

K

King Kazu•10/22/25, 10:30 PM

https://www.inngest.com/docs/local-development

K

King Kazu•10/22/25, 10:30 PM

makes it 10x easier to debug since its nearly impossible with cf workflows rn

KKing Kazu workflows would be great if there was local observability like how inngest has i...

X

xvvvyz•10/23/25, 5:05 AM

yes, please.

D

Deltachaos•10/23/25, 6:56 AM

In Workflows (using Python) i get this kind of exception:
"pyodide.ffi.JsException: Error: Too many API requests by single worker invocation."
Can someone explain to me what API requests are meant? API requests to external APIs using HTTP or API requests to cloudflare services or both?

D

Deltachaos•10/23/25, 10:50 AM

I would love to see a API to get the current limit usage dynamically (within a worker for example the amount of subrequests done)

SShivek 1000 total invocations. - I'm running a workflow that has many steps. - I need...

F

Focux•10/23/25, 1:31 PM

I just came to share this experience and found many comments saying the same. I hope they remove this limit for workers and add more built-in features, such as rate limiting, mechanisms to handle backpressure, child workflows, etc.

F

Focux•10/23/25, 1:31 PM

A

andrew•10/23/25, 7:02 PM

hello, we're heavy users of Workflows and our system is hitting up against the 100 per 10 seconds "Maximum Workflow instance creation rate" limit. we are seeing the number of queued workflow instances as we are adding users.

we've submitted the Limit Increase Request Form twice and wanted to see if someone could help with increasing our limit there

X

xvvvyz•10/23/25, 10:40 PM

need some help with this error if anyone can see something I can't

✘ [ERROR] TypeError: Cannot read properties of undefined (reading 'Symbol(Symbol.dispose)')

      at InstanceImpl.sendEvent
  (file:///Users/cade/Projects/api/node_modules/miniflare/dist/src/workers/workflows/wrapped-binding.worker.js:57:36)

✘ [ERROR] TypeError: Cannot read properties of undefined (reading 'Symbol(Symbol.dispose)')

      at InstanceImpl.sendEvent
  (file:///Users/cade/Projects/api/node_modules/miniflare/dist/src/workers/workflows/wrapped-binding.worker.js:57:36)

✘ [ERROR] TypeError: Cannot read properties of undefined (reading 'Symbol(Symbol.dispose)')

      at InstanceImpl.sendEvent
  (file:///Users/cade/Projects/api/node_modules/miniflare/dist/src/workers/workflows/wrapped-binding.worker.js:57:36)

✘ [ERROR] TypeError: Cannot read properties of undefined (reading 'Symbol(Symbol.dispose)')

      at InstanceImpl.sendEvent
  (file:///Users/cade/Projects/api/node_modules/miniflare/dist/src/workers/workflows/wrapped-binding.worker.js:57:36)

hono endpoint that triggers it

app.post(
  "/webhooks/modal/document-parsed",
  zValidator(
    "json",
    z.object({
      documentId: z.string(),
      instanceId: z.string(),
      markdown: z.string(),
      siteId: z.string(),
    }),
  ),
  async (c) => {
    const { documentId, instanceId, markdown, siteId } = c.req.valid("json");
    const instance = await c.env.DOCUMENT_WORKFLOW.get(instanceId);
    const key = `sites/${siteId}/files/${documentId}.md`;

    await c.env.R2.put(key, markdown, {
      httpMetadata: { contentType: "text/markdown" },
    });

    await instance.sendEvent({
      type: WorkflowEventType.DocumentParsed,
      payload: { markdownFileKey: key },
    });

    return c.json({ status: "ok" });
  },
);

app.post(
  "/webhooks/modal/document-parsed",
  zValidator(
    "json",
    z.object({
      documentId: z.string(),
      instanceId: z.string(),
      markdown: z.string(),
      siteId: z.string(),
    }),
  ),
  async (c) => {
    const { documentId, instanceId, markdown, siteId } = c.req.valid("json");
    const instance = await c.env.DOCUMENT_WORKFLOW.get(instanceId);
    const key = `sites/${siteId}/files/${documentId}.md`;

    await c.env.R2.put(key, markdown, {
      httpMetadata: { contentType: "text/markdown" },
    });

    await instance.sendEvent({
      type: WorkflowEventType.DocumentParsed,
      payload: { markdownFileKey: key },
    });

    return c.json({ status: "ok" });
  },
);

app.post(
  "/webhooks/modal/document-parsed",
  zValidator(
    "json",
    z.object({
      documentId: z.string(),
      instanceId: z.string(),
      markdown: z.string(),
      siteId: z.string(),
    }),
  ),
  async (c) => {
    const { documentId, instanceId, markdown, siteId } = c.req.valid("json");
    const instance = await c.env.DOCUMENT_WORKFLOW.get(instanceId);
    const key = `sites/${siteId}/files/${documentId}.md`;

    await c.env.R2.put(key, markdown, {
      httpMetadata: { contentType: "text/markdown" },
    });

    await instance.sendEvent({
      type: WorkflowEventType.DocumentParsed,
      payload: { markdownFileKey: key },
    });

    return c.json({ status: "ok" });
  },
);

app.post(
  "/webhooks/modal/document-parsed",
  zValidator(
    "json",
    z.object({
      documentId: z.string(),
      instanceId: z.string(),
      markdown: z.string(),
      siteId: z.string(),
    }),
  ),
  async (c) => {
    const { documentId, instanceId, markdown, siteId } = c.req.valid("json");
    const instance = await c.env.DOCUMENT_WORKFLOW.get(instanceId);
    const key = `sites/${siteId}/files/${documentId}.md`;

    await c.env.R2.put(key, markdown, {
      httpMetadata: { contentType: "text/markdown" },
    });

    await instance.sendEvent({
      type: WorkflowEventType.DocumentParsed,
      payload: { markdownFileKey: key },
    });

    return c.json({ status: "ok" });
  },
);

B

ByteBeast•10/24/25, 6:28 AM

Hello, I need one help. I have one workflow with some do steps, if any step got failed and it's maximum retry also got failed so can i retry it manually by any way? cloudflare doesn't provide API to retry specific step manually but by doing any hack can I achieve manual retry for any step?

D

D Trombett•10/25/25, 9:46 AM

Hey not sure if this is a bug or it's expected: Ruleset Engine docs specify that we can use

cf.worker.upstream_zone

cf.worker.upstream_zone

cf.worker.upstream_zone

cf.worker.upstream_zone to identify if the request comes from a Worker but I noticed that this value is empty when the request is from a Workflow. I'm using this to allow requests to a host only to subrequests from my own workers (and thus Workflows)

D

D Trombett•10/25/25, 10:04 AM

Oh wait after investigating a bit I think that the value is not empty but wrong? Instead of my zone (example.com), like requests from my worker, CF-Worker is set to example.workers.dev which doesn't really make sense since my dev url is disabled

DD Trombett Oh wait after investigating a bit I think that the value is not empty but wrong?...

H

Hello, I’m Allie!•10/25/25, 7:07 PM

IIRC that’s what it is set for all requests from Worker Events that aren’t fetches on a zone

H

Hello, I’m Allie!•10/25/25, 7:07 PM

So that’s the correct value

D

D Trombett•10/25/25, 7:13 PM

Oh so even scheduled workers use workers.dev as

upstream_zone

upstream_zone

upstream_zone

upstream_zone? Well, if it's expected is ok, I can still match them with my workers.dev domain, thanks a lot!

DD Trombett Oh so even scheduled workers use workers.dev as `upstream_zone`? Well, if it's e...

H

Hello, I’m Allie!•10/25/25, 8:01 PM

Yeah, all Workers have some kind of upstream zone set, so that you can filter on them. If there are any that don’t, that would be a bug

H

Hunter Bertoson•10/27/25, 9:32 PM

I see in the docs that a

step.sleep

step.sleep

step.sleep

step.sleep is a max duration of 365 days. Is

waitForEvent

waitForEvent

waitForEvent

waitForEvent also that long? Does it have a default timeout if I do not specify one? Or will it literally just wait until the end of time?

D

Directortime•10/27/25, 9:46 PM

I need a server to be put together also a site running on cloudflar dNS is giving a issue.

DDirectortime I need a server to be put together also a site running on cloudflar dNS is givin...

1

1984 Ford Laser•10/27/25, 10:38 PM

Try in #general-discussions or #general-help this isn't the right channel for help with DNS

HHunter Bertoson I see in the docs that a `step.sleep` is a max duration of 365 days. Is `waitFor...

M

Matt Silverlock•10/28/25, 12:33 PM

cc @mia

DDeltachaos I would love to see a API to get the current limit usage dynamically (within a w...

M

Matt Silverlock•10/28/25, 12:34 PM

We’re planning on dramatically increasing this limit (10k+ or even higher) for all paid users very shortly. Within next 2-3 weeks, if not sooner.

MMatt Silverlock cc @mia

H

Hunter Bertoson•10/28/25, 12:42 PM

For some context.

I’m using the waitForEvent for a webhook. And it could be an unknown amount of time before I get that webhook event. It won’t be a year lol. But just wasn’t sure if it had a default timeout that I needed to be aware of.

What's the reason to have a 1000 api request limit per worker ? -- Context: I built a global rate

Similar Threads

Similar Threads

Similar Threads