What version of `wrangler` are you using? Make sure it’s the latest (4.8.0 or later).
What version of
wrangler are you using? Make sure it’s the latest (4.8.0 or later).
BIs there any sample to create workflow using api, as I am trying to make dynamic workflow where user can select how workflow will processed so does anyone have code sample or CURL or anything which can help me out?
DYeah, but the Stripe webhook that trigger the waitForEvent for a particular workflow instance will by definition fire after the workflow has already started. There is no way to put the instance id into Stripe metadata, because there is no way Stripe will have this information. You can't update stripe metadata before something you don't know will happen will happen. This is the case for nearly every single instance where you would use waitFor. The thing that will trigger the waitFor will fire AFTER the workflow starts or what are you waiting for? So there is no way the external event will ever know what the instance is. Reallly the only way I can see that this can possibly work is that both the Workflow and external event that will eventually trigger the waitForEvent both use a predetermined uniqueID that is considered unique across all workflows. In that way the Workflow instance Id would be created with that uniqueId. But this is quite challenging to get right. In the case of orders, I guess you will have a unique order ID that can possibly be shared by Stripe and the Workflow, but this would have to be created by a 3rd party system, so that Stripe and the Worfklow can both know it.
M- You can set your own instance IDs - you don’t have to create the instance first. For example, the Workflow ID can be the order ID so you have a 1:1 association with those. Or your customer ID, transaction ID, or any other stable identifier that you would otherwise use. Workflows is not unique here.
- and/or the Workflow itself can trigger the order, set the webhook, and then wait for the event
MI don’t think anything about this example code is impossible to achieve here.
M
Cloudflare Docs
You can trigger Workflows both programmatically and via the Workflows APIs, including:
BNo, I am not talking about trigger workflow, I want to create workflow script dynamically using script API.
like below I have make CURL for creating workflow script but getting error
like below I have make CURL for creating workflow script but getting error
Mhttps://developers.cloudflare.com/workers/configuration/multipart-upload-metadata/
- see here. You are also missing the Workflow definition. Otherwise, this isn’t specific to Workflows
Cloudflare Docs
If you're using the Workers Script Upload API or Version Upload API directly, multipart/form-data uploads require you to specify a metadata part. This metadata defines the Worker's configuration in JSON format, analogue to the wrangler.toml file.
FFollowing up on this, have updated Wrangler to the latest version (
4.10) but still getting the same errors for running my tests when introducing the workflow
Sin this case, I think that you need to update the
vitest-pool-workers package
FSavior, that helped. This thing has been moving forward fast!
BOkay. thanks @Matt Silverlock
D"You can set your own instance ID" - OK that is not clear from the docs and examples. I think all the examples I have seen use crypto to create a unique instance. Also, I like the idea of the workflow triggering the actual order. Didn't think of that. Thanks!
FWhat we tend to do is use a shared UUID for various items and prefix with what the consumer is, e.g.,
workflow-${uuid} so you can debug more easily too if/when things go southMWhere can we make this clearer? Documented here: https://developers.cloudflare.com/workflows/build/workers-api/#create
An ID is automatically generated, but a user-provided ID can be specified (up to 64 characters 1). This can be useful when mapping Workflows to users, merchants or other identifiers in your system.We kept the tutorial simpler (because otherwise we get feedback that we’re introducing too much!)
M100% this. Same as you would with Temporal or other async workflow systems.
DYeah, I see it there now. I'm not sure how you can make this clearer. It's easy to miss that. And also, I generally read examples first and learn from that, and don't go back to the docs unless I need something new or have a bug. And all the examples either don't create the instance and let it be done automatically or they use crypto. So that's how I've been doing it. Maybe have an example with an instance ID that is user-provided? Not really sure.
MYeah - maybe we can link out from the guide so it’s mentioned but doesn’t pollute the tutorial / add too much up front
MOr a code comment (or both)
FBoth the workers were using 4.9.1. I just tried updating to the latest (4.10.0) but I still get the same error message
FIs there a way to use the chrome debugger inside the workflow steps? I'm unable to hook into it from the main pool, but the worker and DOs work there. Trying to figure out why I'm not getting error logs in a dedicated class that's being called in the step.
BThe way I'm getting around the limit is popping messages into a queue and then batching 100 messages or every 10 seconds to a workflow. When hitting the ratelimit, the queue will redeliver in the next batch
BAlthough my usecase isn't time sensitive, it's handling user registrations
HWe can retry them, of course — but workflows are meant to be about durable execution. What I was expecting was a way to trigger multiple distinct workflows transactionally with a single request. That way, if the Cloudflare Workflows API goes down or we hit a quota/limit issue, either all workflows start or none do — instead of ending up in a half-started, inconsistent state.
By having to manage retries ourselves, we’re essentially recreating the very problem that workflows are supposed to solve, aren’t we?
By having to manage retries ourselves, we’re essentially recreating the very problem that workflows are supposed to solve, aren’t we?
MHow are you solving this today/elsewhere? What requires these Workflows to be tightly coupled in a way that can’t be solved by using one Workflow vs trying to spread across many?
(To be clear, this doesn’t mean we wouldn’t consider a transactional API, but it’s non-trivial across multiple unrelated Workflows, because that has a performance + complexity cost)
(To be clear, this doesn’t mean we wouldn’t consider a transactional API, but it’s non-trivial across multiple unrelated Workflows, because that has a performance + complexity cost)
JWhat's the best way to use Queues to limit the concurrency of Workflows? Or would it be just polling the workflow status and using setTimeout to sleep on the consumer before ack'ing the message?
JOhh yeah, that might work! I've just discovered
waitForEvent
IIs this released or coming in the future? "When creating a Workflow instance, you can set a shorter state retention period if you do not need to retain state for errored or completed Workflows."
MComing soon.
AHello, is there a subrequest limit on workflows? Trying to work out if it fits my use case which I expect to need to make several thousand subrequests to R2. I was looking at a scheduled worker job but I think that certainly does have a 1000 subrequest limit. The docs seem a little unclear in that there is no explicit limit here but it states "Many limits are inherited... [from Workers]".
Cloudflare Docs
Limits that apply to authoring, deploying, and running Workflows are detailed below.
DYep, the limit is the same as workers
AThanks!
SHow can we disable retries? Is is sufficient to set
We did this and still are still seeing retries in our dashboard.
retries0We did this and still are still seeing retries in our dashboard.
Jyou have to configure it for every step
MWhen we will be able to delete Workflows? 
CHi, isn't it possible to search for a workflow run by ID in the CF Dashboard? Or will that be added soon?
KMy workflow works locally but when deployed I get error status on my steps, but I don't have idea how to check the error message.
wrangler tail is not showing any console / error logs
worker logs doesn't show anything either (only invocation)
same for workflows dahsboard (just show status: error)
wrangler tail is not showing any console / error logs
worker logs doesn't show anything either (only invocation)
same for workflows dahsboard (just show status: error)
KLocally I'm getting:
"✘ [ERROR] An RPC stub was not disposed properly. You must call dispose() on all stubs in order to let the other side know that you are no longer using them. You cannot rely on the garbage collector for this because it may take arbitrarily long before actually collecting unreachable objects. As a shortcut, calling dispose() on the result of an RPC call disposes all stubs within it."
but it runs anyway, maybe that is the problem?
"✘ [ERROR] An RPC stub was not disposed properly. You must call dispose() on all stubs in order to let the other side know that you are no longer using them. You cannot rely on the garbage collector for this because it may take arbitrarily long before actually collecting unreachable objects. As a shortcut, calling dispose() on the result of an RPC call disposes all stubs within it."
but it runs anyway, maybe that is the problem?
AReally need a way to limit max concurrency per workflow level. I want some workflows to run 200 in parallel while others only 10. Additionally it would be amazing if workflows can be a queue consumer directly, e.g. take 10 from queue when those 10 are done, take another set of 10.. Any update what is planned and timeline? Also if anyone has an idea how to implement such think today with maybe a workaround, would love to hear that too!
Kthe problem was delay: 0 on retries
Bis it possible to conditionally wait for an event?
Something like that
if (flag === true) await step.waitForEventSomething like that
Bfor regular steps I can do this:
BWoah -- thank you. That makes a lot of sense.
CHi @cloudflare dev , do you have plans to make it possible to jump to another step in the workflow? This should give us a lot more use cases for workflows.
An example:
An example:
- step 1: sleep
- step 2: call some external system to check status
-- when status <> 'done', go back to step 1
-- else when status 'done', continue - step 3: do some processing
HCould you move steps one and two out of the Workflow?
COf course we can. But this is only a very simple example. It should be nice to have the whole process in one workflow.
We build also a lot of workflows in N8N, there we could easily connect output to a previous node to re-run parts of a workflow, until a count or a status result matches, then continue.
I am aware that it could lead to 'spaghetti-code', just like VB-script GOTO
But that's up to us developers.
I'm also not sure if it is hard to implement for CF, because the result of previous steps are accessible in next steps.
But perhaps there was already some discussion about this in the CF team and they can share their thoughts on this.
We build also a lot of workflows in N8N, there we could easily connect output to a previous node to re-run parts of a workflow, until a count or a status result matches, then continue.
I am aware that it could lead to 'spaghetti-code', just like VB-script GOTO
But that's up to us developers. I'm also not sure if it is hard to implement for CF, because the result of previous steps are accessible in next steps.
But perhaps there was already some discussion about this in the CF team and they can share their thoughts on this.
