Cloudflare Developers•4mo ago

Now I'm running into this as well `(

Now I'm running into this as well (workflow.not_found) Provided Workflow does not exist in production. Anyone on the Cloudflare team know what could be happening? Seems a few other people have run into this.

5 Replies

scook•4mo ago

Our team has made an important finding for this bug. We confirmed that when you manually update an Environment Variable for the worker via the dashboard, which triggers a redeploy, it breaks the worker's service binding to the workflow. When we rolled back (via the dashboard) to the deployment version prior to the Env Var change, the bug was no longer reproducible. @Diogo Ferreira @Matt Silverlock We now have a repro for a really nasty bug that breaks workflows bindings. We also suspect there are other ways to trigger the bug aside updating env vars in the dashboard. Repro Steps: 1. Define a Worker (we're using Typescript) 2. Deploy the Worker to cloudflare 3. Define a new workflow that is bound to the worker, using service bindings like so:

# wrangler.toml
[[workflows]]
name = "my-workflow"
binding = "MY_WORKFLOW"
class_name = "MyWorkflowClass"


// Workflow call (Typescript) via service binding:
const instance = await this.env.MY_WORKFLOW.create({
  params: { ... }
});

# wrangler.toml
[[workflows]]
name = "my-workflow"
binding = "MY_WORKFLOW"
class_name = "MyWorkflowClass"


// Workflow call (Typescript) via service binding:
const instance = await this.env.MY_WORKFLOW.create({
  params: { ... }
});

4. Deploy the Worker to cloudflare. 5. Trigger the code that includes the workflow invocation via the service binding. Observe it is successful. 6. In the Cloudflare dashboard for the worker, navigate to Settings > Variables and Secrets. Make any modification to an environment variable that will cause a redeployment of the Worker. 7. Repeat step 5 (trigger the code that includes the workflow invocation). Expected: Workflow invocation is successful. Actual: Error: (workflow.not_found) Provided Workflow does not exist at callFetcher(cloudflare-internal:workflows-api:24:15) at WorkflowImpl.create(cloudflare-internal:workflows-api:78:24) Note: We have also seen this bug occur on the first initial deployment of the workflow (i.e. step 4 above), so there may be a different or a related bug in that flow as well. We have found that performing a full deployment using wrangler will put the worker in a good state again and fix whatever is breaking from step 6, but that is only temporary until you trigger the bad state again.

Diogo Ferreira•4mo ago

@scook thank you for reporting!

TL•3mo ago

Is there an update on this? I've just hit the same bug

Diogo Ferreira•3mo ago

We are working on it. Sorry for the delay @scook @TL this should be fixed at this moment!

scook•3mo ago

Thanks so much!!

Gaming

Programming

Now I'm running into this as well `(

Did you find this page helpful?