@Luís Duarte 🧑🍳 sorry for the ping, but would really love to know your thoughts on this. It would

H

hmndOP•8/18/25, 6:12 AM

For everyone here annoyed by the cumbersome nature of try catching on every step, particularly when you need to use the result of one step in another, remember that because

step.do

step.do

returns a promise, you can use

.catch()

.catch()

on it!

const validationResult = await step
  .do('validate', async () => {
    // do some logic here
    return { somedata: true };
  })
  .catch((e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  });

const validationResult = await step
  .do('validate', async () => {
    // do some logic here
    return { somedata: true };
  })
  .catch((e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  });

const validationResult = await step
  .do('validate', async () => {
    // do some logic here
    return { somedata: true };
  })
  .catch((e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  });

const validationResult = await step
  .do('validate', async () => {
    // do some logic here
    return { somedata: true };
  })
  .catch((e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  });

Or if you are already using

neverthrow

neverthrow

in your project like I am, you can do this!

const validationResult = await fromPromise(
  step.do('validate', async () => {
    //  do some logic that could throw here
    return { somedata: true };
  }),
  (e) => e,
).match(
  (result) => result,
  (e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  },
);

const validationResult = await fromPromise(
  step.do('validate', async () => {
    //  do some logic that could throw here
    return { somedata: true };
  }),
  (e) => e,
).match(
  (result) => result,
  (e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  },
);

const validationResult = await fromPromise(
  step.do('validate', async () => {
    //  do some logic that could throw here
    return { somedata: true };
  }),
  (e) => e,
).match(
  (result) => result,
  (e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  },
);

const validationResult = await fromPromise(
  step.do('validate', async () => {
    //  do some logic that could throw here
    return { somedata: true };
  }),
  (e) => e,
).match(
  (result) => result,
  (e) => {
    // handle your error here, and possibly re-throw it if you want it to be fatal
  },
);

T

tornado•8/18/25, 11:34 PM

was there any recent change to workflow subrequests limit? I'm suddenly seeing some workflows failing with "too many subrequests", this seem to even effect other steps. For example I have a try catch on every step to make sure I update the task status in my db to failed but that fails too because cloudflare isn't allowing the requests (erroring with "too many subrequests") even in different steps

T

tornado•8/18/25, 11:49 PM

nvm my code had edge case that causes infinit loop, thank good I'm using cloudflare AI gateway and all those repeated requests were cached

E

Eumatio•8/19/25, 9:26 PM

guys there is way to know how many retries a step already did?

Hhmnd @Luís Duarte 🧑🍳 sorry for the ping, but would really love to know your thought...

V

Victor•8/20/25, 6:48 AM

https://github.com/LuisDuarte1/flowcesinha https://www.npmjs.com/package/flowcesinha

options.globalFailureHandler

options.globalFailureHandler

V

Victor•8/20/25, 6:55 AM

I have 2 workflows (since closed beta) that have now accumulated over a million of finished (combo of succeeded, errored, and terminated). The backing DO of the workflow itself seems to really struggle (many times getting timed out) just loading instances on the dashboard. And it seems to be ignoring the retention limit

I still have workflows and their full step details (some had bugged code that caused thousands of retries on steps) from 6 months ago. Could someone help "cull" some?

Y

Yu•8/21/25, 6:01 PM

There are too few restrictions on workflow now, my qps is very high, and each task is a long-term task, will the team update this restriction? Or can I only choose other options?

I

illusion•8/22/25, 1:01 PM

how to test workflows? is it the similar like testing workers using vitest?

Iillusion how to test workflows? is it the similar like testing workers using vitest?

M

Matt Silverlock•8/22/25, 1:35 PM

Yep, vitest!

S

scook•8/22/25, 7:59 PM

I'm getting the following error for my workflows that have a sleep step for a longer period of time. Can someone help me troubleshoot?

Error: Aborting engine: Grace period complete

Error: Aborting engine: Grace period complete

MMatt Silverlock Yep, vitest!

I

illusion•8/23/25, 4:09 AM

Is there any examples I can see for workflows? I can't find any examples in https://developers.cloudflare.com/workers/testing/vitest-integration/recipes/

Cloudflare Docs

Recipes and examples

Examples that demonstrate how to write unit and integration tests with the Workers Vitest integration.

D

djoker•8/23/25, 11:33 AM

where can I find logs from cf workflows?

I

illusion•8/25/25, 5:37 AM

I can't use workflow terminate method in local development.

Error: Not implemented yet

Error: Not implemented yet

B

Brandon•8/26/25, 9:08 PM

Hey all we just added support for python workflows

They are still in beta, but would love to get your thoughts, what kind of templates would you like to see, what would you want to build on top of Python workflows?

Here are the docs: https://developers.cloudflare.com/workflows/python/

Please let me know if you need any additional information to get started building, or if there's any clarity I can help provide.

Happy shipping!

Also — fun fact, some of you who are more avid workflow users with lots of resources should hopefully notice a decent performance boost in the dashboard/api.

A

ALECKS•8/27/25, 3:29 PM

is there an quick / easy way to run a workflow locally real quick, I think I saw that the normal CLI does not work locally. Or do I need to add a fetc() api to the workflow and invoke it that way?

D

Devoto•8/27/25, 10:31 PM

Hi there, so I was checking the:

https://developers.cloudflare.com/workflows/reference/limits/

And noticed it says:

Maximum Workflow instance creation rate 100 per 10 seconds

Is this the

env.MY_WORKFLOW.create()

env.MY_WORKFLOW.create()

call? If so, let's say after user registration, we want to setup billings, welcome emails, etc for the user (our most common use-case currently), with our usual load, we would that limit incredibly quickly, what is the solution for these scenarios?

Cloudflare Docs

Limits

Limits that apply to authoring, deploying, and running Workflows are detailed below.

DDevoto Hi there, so I was checking the: https://developers.cloudflare.com/workflows/re...

P

Peps•8/29/25, 12:37 PM

there's a form below that table you referenced, which can be used to request an increase to the limit

V

vmg•8/29/25, 2:45 PM

does a workflow count as a single worker invocation for the purpose of worker sub requests?

L

Levi•8/29/25, 4:03 PM

I'm retrieving topK from KV in one step, and returning it as state to use in the next step. When I return topK from the step where I retrieve it, the type is number. When I try to use it in the next step where I query a vectorize database, it's a string. So, it seems like it serializes my number as a string between steps. Is that intended behavior?

The vectorize query requires topK be a number and I thought I was passing it a number.

L

Lez•9/1/25, 4:03 PM

Hello everyone, hope all is well and happy Monday!

I'm seeing odd "run" errors in our worker logs, but they're not very descriptive and very hard to fix. Has anyone ever seen these before, or is my built just so bad?

Thanks,
Lez

R

Ray Li•9/1/25, 11:53 PM

hey all -- i'm trying to figure out how to bind workflows to my CF workers on pulumi instead of wranglers

V

Victor•9/2/25, 5:03 PM

100 queued, but can't see any, even when set to 12 months

VVictor 100 queued, but can't see any, even when set to 12 months

V

Victor•9/2/25, 10:43 PM

Even hitting the rest API endpoint (with no date parameter), just queued, it's empty. Also a bunch of "waiting" instances, but every step inside of them is complete

L

Lez•9/3/25, 10:32 AM

Hey guys, I'm seeing a lot of these: outcome: "exception", is this normal behavior for workflows?

C

Code Cowboy•9/4/25, 3:48 PM

A request to the Cloudflare API (/accounts/<id>/workflows/<name>) failed.

  Authentication error [code: 10000]

A request to the Cloudflare API (/accounts/<id>/workflows/<name>) failed.

  Authentication error [code: 10000]

A request to the Cloudflare API (/accounts/<id>/workflows/<name>) failed.

  Authentication error [code: 10000]

Anybody else getting this? I can get past it by

wrangler logout && wrangler login

wrangler logout && wrangler login

a few times but its flaky and odd

I'm also having trouble instantiating any workflows at all

T

TaylorBlock•9/4/25, 4:03 PM

you must have permissions that include Account -> Workflows -> Edit/Read (depending on your action).

CCode Cowboy ```A request to the Cloudflare API (/accounts/<id>/workflows/<name>) failed. ...

T

TaylorBlock•9/4/25, 4:04 PM

in order to use API token, you must do this.

TTaylorBlock you must have permissions that include Account -> Workflows -> Edit/Read (depend...

C

Code Cowboy•9/4/25, 4:13 PM

Im the only owner of the account its just me, I definitely have permissions

its weird that wrangler logout && login fixes it

T

tom•9/5/25, 3:10 AM

I have a step that I'm using to process a CSV file and return the contents. The limit for a returned step value is only 1MiB.

Any ideas on how to get around this? The limit is so low, I assume I'm using the steps incorrectly? I'm parsing a CSV into ~10,000 "contact" objects and returning that from the step.

T

tom•9/5/25, 3:11 AM

The exact error is

Step map csv to contacts-1 output is too large. Maximum allowed size is 1MiB.

Step map csv to contacts-1 output is too large. Maximum allowed size is 1MiB.

Ttom I have a step that I'm using to process a CSV file and return the contents. The ...

J

James•9/5/25, 4:35 AM

save it somewhere between steps, r2, d1, do

J

jd•9/5/25, 8:07 PM

hii, is there a full multi-step workflow example with python api? i see step.do only for js but official github repo only have one step request handling without any DAG building / input & output resolution / futures / orchestration / persistence

J

James•9/5/25, 8:55 PM

What's the expected behaviour if you try to create two workflows with the same instance id in quick succession? When testing this locally with

wrangler dev

wrangler dev

it seems the second request is silently dropped, no error is raised but based on log output I can only observe one instance of the workflow being started.

I'm considering triggering a workflow from a queue and wondering if I need to put in additional protection to prevent duplicate workflow creation requests in the case of a message being delivered more than once by the queue. Can I rely on CF to ensure only one workflow gets triggered per instance id?

M

morpig•9/6/25, 9:23 AM

can't get workflows to run reliably using cron. anyone experiencing the same thing?

M

morpig•9/6/25, 9:28 AM

it's only triggering a single workflow instance and yet it's complaining about having too many reqs

M

morpig•9/6/25, 9:28 AM

JJames What's the expected behaviour if you try to create two workflows with the same i...

M

morpig•9/6/25, 9:47 AM

are you talking about cloudflare queues? afaik there's a mechanism in place to not send duplicate messages to handler/http pull until ACK timeout is hit.

Mmorpig are you talking about cloudflare queues? afaik there's a mechanism in place to n...

J

James•9/6/25, 11:22 AM

Yeah, CF queues. My understanding was it's at least once delivery so there's always a chance of duplicate deliveries from the queue to a consumer regardless of any ack timeouts.

J

James•9/6/25, 12:41 PM

I'm also facing an unrelated issue, I don't seem to be able to call a Durable Object from a Workflow, either inside a step or within the run() function. An error like this is returned and then the workflow stops in the

Errored

Errored

status.

internal error; reference = qkhe7o6ej7li8rrie5v0eks4

internal error; reference = qkhe7o6ej7li8rrie5v0eks4

Interestingly this works fine when I run locally via

wrangler dev

wrangler dev

. This is a simplified version of my workflow/DO:

import { WorkflowEntrypoint, WorkflowStep, WorkflowEvent, DurableObject } from 'cloudflare:workers';
import { ActionEvent } from '../types/events';

export class Portal extends DurableObject<Env> {
  constructor(ctx: DurableObjectState, env: Env) {
    super(ctx, env);

    console.log('Initialising portal Durable Object');
  }

  async test(): string {
    return "Hello!";
  }
}

export class ActionWorkflow extends WorkflowEntrypoint<Env, ActionEvent> {
  async run(params: WorkflowEvent<ActionEvent>, step: WorkflowStep) {
    const event = params.payload;
    const portalIdString = event.portal_id.toString();
    console.log('Starting workflow', event);

    const portal = this.env.PORTAL.getByName(portalIdString);
    const test = await portal.test();

    await step.do('test', async () => {
      console.log(test);
    });
  }
}

import { WorkflowEntrypoint, WorkflowStep, WorkflowEvent, DurableObject } from 'cloudflare:workers';
import { ActionEvent } from '../types/events';

export class Portal extends DurableObject<Env> {
  constructor(ctx: DurableObjectState, env: Env) {
    super(ctx, env);

    console.log('Initialising portal Durable Object');
  }

  async test(): string {
    return "Hello!";
  }
}

export class ActionWorkflow extends WorkflowEntrypoint<Env, ActionEvent> {
  async run(params: WorkflowEvent<ActionEvent>, step: WorkflowStep) {
    const event = params.payload;
    const portalIdString = event.portal_id.toString();
    console.log('Starting workflow', event);

    const portal = this.env.PORTAL.getByName(portalIdString);
    const test = await portal.test();

    await step.do('test', async () => {
      console.log(test);
    });
  }
}

import { WorkflowEntrypoint, WorkflowStep, WorkflowEvent, DurableObject } from 'cloudflare:workers';
import { ActionEvent } from '../types/events';

export class Portal extends DurableObject<Env> {
  constructor(ctx: DurableObjectState, env: Env) {
    super(ctx, env);

    console.log('Initialising portal Durable Object');
  }

  async test(): string {
    return "Hello!";
  }
}

export class ActionWorkflow extends WorkflowEntrypoint<Env, ActionEvent> {
  async run(params: WorkflowEvent<ActionEvent>, step: WorkflowStep) {
    const event = params.payload;
    const portalIdString = event.portal_id.toString();
    console.log('Starting workflow', event);

    const portal = this.env.PORTAL.getByName(portalIdString);
    const test = await portal.test();

    await step.do('test', async () => {
      console.log(test);
    });
  }
}

import { WorkflowEntrypoint, WorkflowStep, WorkflowEvent, DurableObject } from 'cloudflare:workers';
import { ActionEvent } from '../types/events';

export class Portal extends DurableObject<Env> {
  constructor(ctx: DurableObjectState, env: Env) {
    super(ctx, env);

    console.log('Initialising portal Durable Object');
  }

  async test(): string {
    return "Hello!";
  }
}

export class ActionWorkflow extends WorkflowEntrypoint<Env, ActionEvent> {
  async run(params: WorkflowEvent<ActionEvent>, step: WorkflowStep) {
    const event = params.payload;
    const portalIdString = event.portal_id.toString();
    console.log('Starting workflow', event);

    const portal = this.env.PORTAL.getByName(portalIdString);
    const test = await portal.test();

    await step.do('test', async () => {
      console.log(test);
    });
  }
}

Any ideas? Or is anyone from CF able to look up that error reference

JJames I'm also facing an unrelated issue, I don't seem to be able to call a Durable Ob...

A

Anes•9/6/25, 2:45 PM

Workflows only allow you to keep state that’s returned from steps. Any call to a Durable Object (or other async call) has to happen inside a step.do(), otherwise even if it works locally in wrangler dev it fails in production.

Here’s the corrected version:

const portalResponse = await step.do('get-portal-response', async () => {
  const portal = this.env.PORTAL.getByName(portalIdString);
  return await portal.test();
});

const portalResponse = await step.do('get-portal-response', async () => {
  const portal = this.env.PORTAL.getByName(portalIdString);
  return await portal.test();
});

const portalResponse = await step.do('get-portal-response', async () => {
  const portal = this.env.PORTAL.getByName(portalIdString);
  return await portal.test();
});

const portalResponse = await step.do('get-portal-response', async () => {
  const portal = this.env.PORTAL.getByName(portalIdString);
  return await portal.test();
});

Cloudflare Docs

Rules of Workflows

A Workflow contains one or more steps. Each step is a self-contained, individually retriable component of a Workflow. Steps may emit (optional) state that allows a Workflow to persist and continue from that step, even if a Workflow fails due to a network or infrastructure issue.

R

rbravo•9/6/25, 9:14 PM

Hey, Cloudflare team! I'm heavily using Cloudflare Workflows for one of the core features of my company, and I'm sometimes reaching the limits, which makes this feature unreliable. I've filled out the Limit Increase Request Form but never heard back—can someone help me? I need to increase some of our limits by 10x.

AAnes Workflows only allow you to keep state that’s returned from steps. Any call to a...

J

James•9/7/25, 7:31 AM

Thanks for the reply, that is initially how I had my workflow set up to call the DO from inside a step but it still fails like that also.

I tried your version exactly and still get an error response from the step, e.g:

internal error; reference = 4t7lk8kksul8hag46prmnndt

internal error; reference = 4t7lk8kksul8hag46prmnndt

J

James•9/7/25, 7:53 AM

Something must be messed up with the state of one of my DO instances, I tried the exact same code but with a different instance ID and it works fine now

Rrbravo Hey, Cloudflare team! I'm heavily using Cloudflare Workflows for one of the core...

M

Matt Silverlock•9/8/25, 4:16 PM

Which limits? Not all of the limits can be increased (many are real, hard limits).

Rrbravo Hey, Cloudflare team! I'm heavily using Cloudflare Workflows for one of the core...

M

Matt Silverlock•9/8/25, 4:32 PM

cc @mia for vis

Mmorpig can't get workflows to run reliably using cron. anyone experiencing the same thi...

M

morpig•9/8/25, 4:52 PM

workflows with cron is still unreliable for us. hope the team knows about this issue.

M

morpig•9/8/25, 4:54 PM

wondering if we're triggering it the wrong way

R

rbravo•9/8/25, 11:40 PM

@Matt @mia here's my use case for more context:

Let's say that I want to send a marketing campaign with messages for 10k+ users. And I'd like these messases to have a random interval between then within a fixed limit (e.g 10s to 120s).

The way I'm creating this now is that for each message I schedule a workflow run.

In the docs it's mentioned that "Concurrent Workflow instances (executions) per account" has a limit for 4500, I believe this is the main one that I need to be increased.

For workflows creation and deletion (let's say that the user has pressed a "Cancel campaign" button) I'm receiving a bunch of 429, but that probably a skilll issue on my part and an exponential backoff should fix it...

I'd appreciate any recs. Thank you!

@Luís Duarte 🧑🍳 sorry for the ping, but would really love to know your thoughts on this. It would

Similar Threads

Similar Threads

Similar Threads