Howdy all, I'm trying about workflow and saw this doc to running it locally, https://developers.clou

Howdy all, I'm trying about workflow and saw this doc to running it locally, https://developers.cloudflare.com/workflows/build/local-development/, is there a way for me to test out the workflow? I know that

npx wrangler workflow

npx wrangler workflow

targets production api. Do I need to setup a local worker to call my local workflow?

Cloudflare Docs

Local Development · Cloudflare Workflows docs

Workflows support local development using Wrangler, the command-line interface for Workers. Wrangler runs an emulated version of Workflows compared to the one that Cloudflare runs globally.

WWilson Wong Howdy all, I'm trying about workflow and saw this doc to running it locally, htt...

Matt Silverlock•12/20/24, 1:17 PM

The doc you linked shows the steps: https://developers.cloudflare.com/workflows/build/local-development/#start-a-local-development-session

Cloudflare Docs

Local Development · Cloudflare Workflows docs

Workflows support local development using Wrangler, the command-line interface for Workers. Wrangler runs an emulated version of Workflows compared to the one that Cloudflare runs globally.

Fernando Barroso•12/20/24, 9:25 PM

Is there a way to delete workflows on our accounts? I can’t seem to find it either in the dashboard or api

Fernando Barroso•12/20/24, 10:28 PM

Oops! What happens if you delete the worker bound to a workflow? Can that workflow still be used by another worker?

hkdobrev•12/21/24, 6:45 AM

I'm using workflows and I really like the built-in design to retry individual steps, very much like a multi-queue solution for each step.

However, I'm also using queues to trigger N instances of the workflow as I want to space out some external calls, so I'm using delay with the queue.
But I'd very much like the workflow to receive the original batch.
This would allow for a much more robust acknowledgement and retry mechanism. If the workflow has access to the JS API of the queue batch, it will be able to control acknowledgment of the message. By retrying the queue message, it will be able to revert back executed steps by retrying the queue message and triggering a new workfow instance resetting the state.
The step sleep is not enough for that.

Currently, I need to resort to acknowledging all messages in the queue batch as soon as the workflow is triggered regarless if it's successful.

SSeekerdasbatatas update your `@cloudflare/worker-types` and should be on `WorkflowEvent.instanceI...

alex (he/him)•12/21/24, 7:06 AM

Doesn't seem to work with

wrangler types --experimental-include-runtime

wrangler types --experimental-include-runtime

D Trombett•12/21/24, 11:08 AM

Assuming that I have a workflow which at the end returns something, can I later retrieve that value using

env.WORKFLOW.get(id).status().output

env.WORKFLOW.get(id).status().output

(something like a sort of kv?) or is it something one shouldn't rely on (e.g. expires after some time)?

D Trombett•12/21/24, 11:35 AM

Ah, I guess this is part of the following limit:

Retention limit for completed Workflow state: 3 days; 30 days

Ashley•12/21/24, 6:11 PM

Just as an FYI, I think there's some out of date info in the docs referring to

event.params

event.params

(e.g. https://developers.cloudflare.com/workflows/get-started/guide/) whereas I now think it's

event.payload

event.payload

(changing this fixed it for me so guessing event.params was old?)

Warren•12/22/24, 4:35 AM

Asking a question re the

Serialized RPC arguments or return values are limited to 1MiB

Serialized RPC arguments or return values are limited to 1MiB

error

the limits:

maximum persisted state per step is 1MiB
Workers return value can only be 1MiB

as such, what I'm trying to do is to ensure that the work done in step in is smaller than the 1MiB limit before returning it.

I am using this to get the size of the return value

export function getApproximateSizeInBytes(obj: unknown) {
  return new TextEncoder().encode(JSON.stringify(obj)).length;
}

export function getApproximateSizeInBytes(obj: unknown) {
  return new TextEncoder().encode(JSON.stringify(obj)).length;
}

however, I notice that the value returned and the actual value shown when an error is thrown can be quite different.

how can I reliably check what is the size of a return value before returning it?

(guess this is not specific to Workflows, but it's relevant in this context because a workaround I've seen is to use ReadableStream, but that wouldn't work because of the max persisted state per step limit)

BBenQoder Do pending workflows run on newly deployed code? Or will it run on the code tha...

Matt Silverlock•12/24/24, 12:07 AM

Always the code as it was deployed. We’re thinking about how to enable some options here. Lots of divisive opinions.

BruceX•12/25/24, 9:24 PM

Hey guys is it not possible to remove a deployed workflow?

eduardo | Latitude.sh•12/26/24, 12:16 PM

Hi everyone, any luck setting up Bugsnag?

oof2win2•12/26/24, 3:01 PM

is it possible to increase the number of maximum outbound requests, higher than 1k? splitting the workflow into multiple executions probably is a possibility but it seems kinda weird as a limit

yzziK•12/26/24, 9:43 PM

agreed ^

yzziK•12/27/24, 1:15 PM

hi will the

step_name

step_name

be added anywhere in future in status ?
like the cloudflare dashboard shows the steps status so its easier to track on which step is currently processing.

i return

step_name

step_name

along with the data as a workaround

ᴀ

ᴀkqj•12/28/24, 1:54 AM

Hey team, is there an ETA on when the concurrent execution limit of 100 will be bumped? would the workflows in await steps count toward the concurrent execution limit?
we have workflows which have a total execution time of less than 10 m and want to be able to run 50K per minute and it is not a constant rate.

ᴀᴀkqj Hey team, is there an ETA on when the concurrent execution limit of 100 will be ...

Matt Silverlock•12/28/24, 2:31 AM

Not yet. Are you asking for ~833/sec or much higher? Have you tried using a Queue to manage a backlog / back off?

ᴀ

ᴀkqj•12/28/24, 3:10 AM

yes 900/s or more concurrent WF, right now we are using the queue and are event based, liked the workflow model which fits our use case, we want to have multiple campaign workflows which would create multiple subscribers workflow. subscribe workflow will perform bunch steps and send out the content

ᴀᴀkqj yes 900/s or more concurrent WF, right now we are using the queue and are event ...

Matt Silverlock•12/28/24, 5:43 PM

We’re still early-ish in the beta so won’t be 900-1000+ concurrent overnight. Recommend using a Queue here - even at scale, a Queue is going to give you the mechanisms to manage the backlog

Carl•12/28/24, 7:55 PM

any update on the infinite queued instances dilemma. currently, i'm unable to start a new instance of any workflow account wide.

CCarl any update on the infinite queued instances dilemma. currently, i'm unable to st...

Matt Silverlock•12/28/24, 10:18 PM

This should have been fixed last week. Can you share your account ID so we can see what’s off here?

MMatt Silverlock This should have been fixed last week. Can you share your account ID so we can s...

Carl•12/29/24, 8:07 AM

e62618c17826d22a8c7630187706413b

CCarl e62618c17826d22a8c7630187706413b

Matt Silverlock•12/30/24, 11:22 AM

This should be fixed now: we believe we fixed the root cause on this (metadata out of sync).

MMatt Silverlock This should be fixed now: we believe we fixed the root cause on this (metadata o...

Carl•12/30/24, 12:40 PM

unfortunately still facing the issue. both for existing workflows and a new basic example one i created to test.

CCarl unfortunately still facing the issue. both for existing workflows and a new basi...

Matt Silverlock•12/30/24, 2:17 PM

@Diogo Ferreira is looking into this now

felixv•12/30/24, 11:34 PM

Is there a way to create a workflow and wait until it’s finished ?

what would be the most elegant way.

eg. I have a hono api. Then I wanna call a workflow and once it’s finished return the result.

Is polling via promise resolve the best ?

Ffelixv Is there a way to create a workflow and wait until it’s finished ? 🤔 what would...

tadhglewis•12/31/24, 12:02 AM

It doesn't seem like a good fit to me. You're trying to do asynchronous work in a synchronous flow. Why not just do your workflow logic in the hono API?

tadhglewis•12/31/24, 12:04 AM

The entire thing with Workflows is it doesn't provide guarantees on when it will finish or how quickly. You would run into issues with connection timeouts etc if you try to keep an API connection open

tadhglewis•12/31/24, 12:05 AM

It's hard without knowledge of what you're doing but a common standard if you need to notify a consumer once async work is done is by sending them a webhook

jrydberg•12/31/24, 12:23 AM

Any plans on increasing the step result size limit from 1MB ?

felixv•12/31/24, 7:24 AM

Awesome! Thanks

. That makes a lot of sense.

tadhglewis•1/1/25, 2:11 AM

Are there plans to make Workflows recoverable? As in, previously I was using a queue + worker and failed items would go to the dead letter queue, be manually investigated and redrived/replayed if there was a code bug.

Will it be possible to retry "dead letter" Workflows?

Ttadhglewis Are there plans to make Workflows recoverable? As in, previously I was using a q...

Matt Silverlock•1/1/25, 2:03 PM

How would you want this to work? Replay the instance (which would make tracking by ID complicated) with the same params?

I think this is tied up in how we tackle versioning, which is something we’re working on but non-trivial. Some folks want workflow instances to be immutable; others for them to reflect code updates; and a few want them to update between steps (unlikely).

I could imagine a

.create({fromExisting: <id>})

.create({fromExisting: <id>})

or similar that allows you to spawn a NEW instance using the input params of the old.

progame•1/1/25, 6:04 PM

what

jrydberg•1/1/25, 8:32 PM

I've started playing with workflows, and ran into the following issue. wonder if it could be related to Workflows?

https://discordapp.com/channels/595317990191398933/1324112336121888858/1324112336121888858

MMatt Silverlock How would you want this to work? Replay the instance (which would make tracking ...

tadhglewis•1/1/25, 11:29 PM

So for context, my workflow does something like

Creates a new workflow instance with an account id which needs creating/updating
Fetch data from internal db and create/update data in a 3rd party via their API

I need every workflow to have always passed (and if one fails, send an alert via our monitoring). I am also planning on surfacing the workflow status to the end consumer by adding another step which updates our internal DB to reference the workflow instance.

Having said that, for my use case, just creating a new instance with the same params would work. I suppose I could do this already using the Workflows API

Ttadhglewis So for context, my workflow does something like 1. Creates a new workflow insta...

Matt Silverlock•1/2/25, 1:37 PM

Right. We could make that (duplication) easier, but if the previous instance’s data/state expires, it’d be more error prone.

Ttadhglewis It doesn't seem like a good fit to me. You're trying to do asynchronous work in ...

Focux•1/2/25, 5:45 PM

I disagree with this. Sometimes, for some tasks, you don't want the async nature of the workflow but the durable and retry functionality without having to implement it yourself.

Focux•1/2/25, 5:49 PM

Some workflows engine already supports that, like Restate, for example.

FFocux I disagree with this. Sometimes, for some tasks, you don't want the async nature...

tadhglewis•1/2/25, 9:38 PM

IMO sounds like bad practice. If you want retries, configure that in your HTTP client (Axios, ky etc).

I've never heard of Restate and it doesn't look like it's being used at scale.

Ttadhglewis IMO sounds like bad practice. If you want retries, configure that in your HTTP c...

Focux•1/2/25, 9:39 PM

You lost me at "it doesn't look like it's being used at scale", end of discussion lol

Hugi•1/2/25, 9:39 PM

@isaacmcfadyen @isaacmcfadyen

tadhglewis•1/2/25, 9:40 PM

Anyhow I was specifically talking about CloudFlare Workflows and how they work (asynchronously)

Focux•1/2/25, 9:41 PM

And I was talking about durable workflows and how the only upside is not async execution

Focux•1/2/25, 9:42 PM

If that's the case, there are better-suited tools for that use case

andyy•1/3/25, 11:09 AM

I'd like to be able to treat a workflow parameter as secret similar to how workers secrets are treated - i.e. value only retrievable within the workflow execution itself (vs. currently parameter values are output in workflow dashboard/logs etc). doesn't seem possible currently but seemed worth mentioning. use case is to pass in short-lived external service authentication credentials to a workflow run that are unique to each user in a multi-tenant system (perhaps there's a better way to implement this). thanks!

Aandyy I'd like to be able to treat a workflow parameter as secret similar to how worke...

tadhglewis•1/3/25, 12:10 PM

I'd assume these credentials are stored somewhere in your application database?

tadhglewis•1/3/25, 12:11 PM

You'd probably be better off passing in a key/ID which correlates to Secrets Manager (or a database row) which contains the creds

Howdy all, I'm trying about workflow and saw this doc to running it locally, https://developers.clou

Similar Threads