Durable Object alarms are better suited for this

Yash Gupta•7/21/25, 9:29 AM

Has anyone tried to access steps/execution history for workflows from within their application?

It seems that the workers binding for workflows only gives present status and output, and while the API itself has steps history, it doesn't work for local development, making it hard to incorporate that as part of the solution.

Another option I've explored is manually intrumenting the step, waitForEvent and sleep methods, but it seems like redundant work (and possibly flaky)

alex (he/him)•7/22/25, 11:44 AM

I'm really having some issues with workflow creations

alex (he/him)•7/22/25, 11:44 AM

Already using

createBatch

createBatch

and try to create ~650 instances all at the same time

alex (he/him)•7/22/25, 11:45 AM

That would be 7 calls with

createBatch

createBatch

.. however I'm still seeing super long time until

createBatch

createBatch

resolves as well as a lot of too many requeststoo many requests errors.

alex (he/him)•7/22/25, 11:45 AM

This is super hard to debug and annoying too.. why does even

createBatch

createBatch

limits soo hard?

Bboop Durable Object alarms are better suited for this

Rahul gangotri•7/22/25, 1:09 PM

alarms will be expensive as it also charges on wall time, so if a task is taking longer to resolve or making call to external api is in place, then its going to cost for all that wait time.

ac•7/22/25, 2:08 PM

How likely is it that the "Maximum Workflow instance creation rate (100 per 10 seconds)" limit will be increased? This limit is pretty low for us

RRahul gangotri alarms will be expensive as it also charges on wall time, so if a task is taking...

boopOP•7/22/25, 2:20 PM

Make the DO call a workflow, just use the DO alarms as a scheduler

Aac How likely is it that the "Maximum Workflow instance creation rate (100 per 10 s...

alex (he/him)•7/22/25, 3:00 PM

This limit is pretty low for us

It's the same for us. We are already looking at alternatives to Cloudflare Workflows, since it doesn't work for high load use cases.

Aalex (he/him)> This limit is pretty low for us It's the same for us. We are already looking ...

Jeffrey Konowitch•7/22/25, 3:46 PM

I've asked about this three times and no response from Cloudflare. A bit concerning for anyone looking to build apps with even moderate throughput requirements.

workflows
Clarifying `Maximum Workflow instance creation rate` Limit
Is the workflow creation rate limit 100 per 10 seconds for createBatch?

hype•7/22/25, 4:07 PM

Same. Just running into this when trying to migrate

hype•7/22/25, 4:07 PM

and now looking again into alternatives.

Hhype and now looking again into alternatives.

alex (he/him)•7/22/25, 4:32 PM

There are not too many for our usage pattern. Also noone can really beat Cloudflare's pricing.

JJeffrey Konowitch I've asked about this three times and no response from Cloudflare. A bit concern...

alex (he/him)•7/22/25, 4:34 PM

I think what could work is use Cloudflare Queues and then consume settings to take batches of 100 each and use

createBatch

createBatch

, but then we would end up with

QUEUE A -> WORKFLOW A -> QUEUE B -> WORKFLOW B

QUEUE A -> WORKFLOW A -> QUEUE B -> WORKFLOW B

at which point I would argue, we can rely on Queues solely and not use Workflows at all.

alex (he/him)•7/22/25, 4:37 PM

So the fact that we also can't control concurrency on a per Workflow level means, that if you publish 600 Instances using

createBatch

createBatch

Workflow A

Workflow A

and every instance uses

create

create

to run

Workflow B

Workflow B

that will not work, because as soon as 101 or more of

Workflow A

Workflow A

run in parallel, we already hit the creation limit to create those "child" workflows

alex (he/him)•7/22/25, 4:38 PM

@Walshy

this would be Feedback you asked for on X

Walshy•7/22/25, 4:39 PM

cc: @Sid Chatterjee @Diogo Ferreira

Aalex (he/him)> This limit is pretty low for us It's the same for us. We are already looking ...

James•7/22/25, 5:51 PM

this was the main reason we didn't implement workflows, it wasn't much work to make something similar out of DOs and Queues that suits the same purpose and stateful guarantees.

Bboop Make the DO call a workflow, just use the DO alarms as a scheduler

Rahul gangotri•7/22/25, 6:11 PM

workflow has very weird bugs right now, all of my instances got queue and never got released, can't trust it right now

JJames this was the main reason we didn't implement workflows, it wasn't much work to m...

alex (he/him)•7/22/25, 6:29 PM

I would be interested how you use DOs to make something similar to Workflows

RRahul gangotri workflow has very weird bugs right now, all of my instances got queue and never ...

boopOP•7/22/25, 6:32 PM

I'm not too sure what this has to do with my suggestion

Elliot Hesp•7/22/25, 9:56 PM

Are there plans to get restart/pause etc working in local development?

alex (he/him)•7/23/25, 6:56 AM

And the most annoying part is actually, that running things locally with

wrangler dev

wrangler dev

does not have any limits in place, that means we can develop something which looks like it would work, but then it doesn't work at all in production, due to those limits only applying in production.

Vlad Lykhonis•7/23/25, 10:17 AM

It is been some time since I have posted regarding longer wall times for workflows. It seems quite random and eventually recovers, but consider my steps consist of simple few REST calls, it seems quite slow. Here is my latest instance for reference. I am not quite sure how to dig into this further to understand the possible cause of this.

Vlad Lykhonis•7/23/25, 10:19 AM

and right after it, I just ran more steps, and it was fast!

Aalex (he/him)So the fact that we also can't control concurrency on a per Workflow level means...

Vlad Lykhonis•7/23/25, 10:28 AM

Naturally I had the same first thought process - create many workflows, each workflow is small unit of work sequence. Then I saw limits in docs. Now, I rather pass batched arguments to fewer workflows and each workflow does a loop of steps per each unit of work sequence.

In the end it is not much of rework, and keeps it all still within limits and achiving same results.

alex (he/him)•7/23/25, 10:34 AM

We will also hit the step limit of 1024 if we start do run more work in one instance.

Aalex (he/him)> This limit is pretty low for us It's the same for us. We are already looking ...

ddsgadget•7/23/25, 1:18 PM

check out upstash workflows. I'm kind of surprised Cloudflare hasn't acquired Upstash. For some stuff we use Cloudflare Workflows, but the UI and for complex cases we use upstash now.

Elliot Hesp•7/23/25, 3:39 PM

What are the rules for a unique workflow ID? Trying my own but I get

invalid_id

invalid_id

no idea if it's the characters or length, but it works locally with whatever

WWalshy cc: @Sid Chatterjee @Diogo Ferreira

Jeffrey Konowitch•7/25/25, 6:58 PM

FYI still waiting on clarification from Cloudflare:

workflows

JJeffrey Konowitch FYI still waiting on clarification from Cloudflare: https://discord.com/channel...

dandoen•7/25/25, 10:27 PM

also interested in this. huge impact on design decisions (and frankly whether workflow fits many use cases or we need to stick to good ol' temporal)

Tory•7/26/25, 5:27 PM

We started seeing this error yesterday in one of our workflows

The RPC receiver does not implement the method "run"

The RPC receiver does not implement the method "run"

. It seems to happen randomly after a successful build of the worker. The only thing that fixes it is re-running the build.

Has anyone else seen this recently?

surgery18•7/28/25, 8:20 PM

I have a step in a workflow that could take multiple minutes to run (Google Gemini API call generating TTS of a script), it keeps hitting 100 seconds and then errors out. Any way to fix that?

surgery18•7/28/25, 8:25 PM

The step returns this

ApiError: {"error":{"message":"error code: 524","code":524,"status":""}}

ApiError: {"error":{"message":"error code: 524","code":524,"status":""}}

which is a timeout error. I am guessing steps have a limit of 100 seconds and then kills the network call. Any way to have the step last as long as the network call lasts?

Ssurgery18 The step returns this `ApiError: {"error":{"message":"error code: 524","code":52...

Chaika•7/29/25, 12:11 AM

That sounds more like the generic proxy limit then Workflows itself.
Cloudflare as a cdn/proxy has a 100 second read timeout which will respond with a 524 if you exceed it: https://developers.cloudflare.com/support/troubleshooting/http-status-codes/cloudflare-5xx-errors/error-524/.
If the body starts responding/streaming you can get around this, or maybe the Gemini API has a different way to poll for updates. Normally you can also raise this if you are Enterprise as well but since Workflows don't run on a specific website, might be difficult.

surgery18•7/29/25, 12:15 AM

Hmmm...Ok. I will see if Gemini API has a streaming option. Probably does. I will try that then. Thanks.

JJeffrey Konowitch FYI still waiting on clarification from Cloudflare: https://discord.com/channel...

Brandon•7/29/25, 4:27 PM

Hey apologies on the delay here, we are actively working on upping the concurrency limits on workflows, this will be shipping in the near future and I can provide more details as we get closer to going live with it.

hype•7/31/25, 11:10 AM

Do workflows also share same connection limits that DO/worker has? As in you cannot do more than 6 fetch calls simultaneously?

Developres•7/31/25, 11:11 AM

Hello everyone, I want to use cdn. I haven't used it on my website before. How do I use it? Can you send me a video?

hype•7/31/25, 7:50 PM

Is there a limit on how long a workflow can run?

hype•7/31/25, 7:50 PM

Can it have infinitely long running task on it? How long each step can run for?

Cristian Lipinski•8/1/25, 10:16 AM

Hello guys,

since yesterday (2025-07-31) my workflow instances started by "createBatch" are not using my defined id, as the documentation says:

// Create a new batch of 3 Workflow instances, each with its own ID and pass params to the Workflow instances
const listOfInstances = [
{ id: "id-abc123", params: { "hello": "world-0" } },
{ id: "id-def456", params: { "hello": "world-1" } },
{ id: "id-ghi789", params: { "hello": "world-2" } }
];
let instances = await env.MY_WORKFLOW.createBatch(listOfInstances);

its creating its own id as if we were creating it without passing the parameter, we have dozens of workflows so it does not seen like a isolated thing

could anyone help me with that? thanks

Cristian Lipinski•8/1/25, 10:29 AM

I would also like to report another problem that we're seeing in some workflow instances, our account has multiple domains, we have some workflows that make some on more than a hundred domains once by day

some of my instances, although it completes all steps, are returning "with error" status, for example, the instance ID: 0ec749d5-fdb9-4dbd-b8da-b1a03bf4ca2e

error description:
Durable Object storage operation exceeded timeout which caused object to be reset.

CCristian Lipinski I would also like to report another problem that we're seeing in some workflow i...

avenceslau•8/1/25, 10:31 AM

Will also investigate that one

toxec•8/1/25, 5:58 PM

Hello, what’s the easiest way to manually trigger a workflow in a local environment? I know I can’t just use wrangler, so my current workaround is to set up local-only routes and invoke them with curl, but it doesn’t feel ideal. I’m using Hono with Vite.

Cristian Lipinski•8/4/25, 1:21 PM

Hello there guys.

We're passing though really tough loading problems on my companies main account, all our GUI is way too slow, im uploading some examples on what we are experiencing

we work with 4 cf accounts, and only on our main account, that has hundreds of workflows, workers, kv, d1 is passing through all those loading problems

beside the loading, we're also havind a hard time deploying our workers, im also uploading examples

lastly, some of our workflows, even using "createBatch" are failing, with "Internal error"

What should we do about that? Maybe spliting our routines on other accounts? Is there anything to do in order to increase our performance?

Please let me know if there's anything we could do, we really like most of CF services, but as we're scaling, its not working as we expect.

we're even considering options like trigger.dev, its really frustrating because we really love CF, but if we can't trust it on high numbers,its hard to keep working with it

fforres•8/4/25, 9:38 PM

Hey folks

Saw that the column sorting on the workflow list page is not working properly.

Seems to sort by some logic, but not by date value or retries, etc.

fforres•8/4/25, 9:39 PM

here by retries or cpu time

Brandon•8/4/25, 11:21 PM

I'll look into this, thank you for flagging!

Victor•8/5/25, 1:20 AM

Getting

workflows.api.error.internal_server [code: 10001]

workflows.api.error.internal_server [code: 10001]

wrangler workflows ...

wrangler workflows ...

wrangler publish

wrangler publish

, web dashboard, and even normal api calls

Durable Object alarms are better suited for this

Similar Threads

Similar Threads

Similar Threads