Regression since 2.27.0: waiting webhook resume fails in queue mode when parent execution is not saved

Questions
1

Title:
Regression since 2.27.0: waiting webhook resume fails in queue mode when parent execution is not saved

Hi n8n community,

We are blocked from upgrading our self-hosted production instance beyond 2.26.9 because of a regression introduced between 2.26.9 and 2.27.0.

GitHub issue:

Setup:

  • self-hosted n8n
  • queue mode
  • one main instance + workers
  • affected flow uses Wait / webhook-waiting resume through a sub-workflow flow
  • affected workflows have saveDataSuccessExecution: none

Observed behavior:

  • works on 2.26.9
  • fails on 2.27.0
  • still fails on 2.28.0
  • we have not found anything in the 2.28.x / 2.29.x release notes indicating this was fixed

Failure:
NODE FAILED: Call webhook
ERROR: The resource you are requesting could not be found

We narrowed the regression window to between 2.26.9 and 2.27.0.
The strongest candidate PR seems to be:

Reason: it changes the parent/child resume path after a Wait and calls updateParentExecutionWithChildResults(...) before resuming the parent execution.

Has anyone else reproduced this with queue mode + Wait/webhook-waiting + saveDataSuccessExecution: none?

Any suggested workaround other than staying pinned to 2.26.9?

2

Hi @zlebandit

Can you try this?

  1. Open the Parent Workflow settings.
  2. Change the Save successful production executions setting from “Do not save” (none) to “Save all” (all) or “Save latest” (latest).
  3. Save the workflow and redeploy.

This prevents the execution from being soft-deleted while parked at the Wait node, allowing the resume webhook to find the execution and continue the flow.

3

Thanks for the suggestion.

The problem is precisely that we do not want to save all successful executions.

This instance runs many production workflows, and saveDataSuccessExecution: none is intentional: we want to avoid filling the executions list and database with successful runs that have no operational value once completed.

Also, this setup worked correctly on 2.26.9 with the same workflow settings. The regression starts in 2.27.0.

So enabling successful execution saving may work around the symptom, but it is not really acceptable as a long-term fix. A waiting execution should keep whatever minimal resume state it needs until the Wait/webhook-waiting flow completes, without requiring all successful executions to be saved permanently.

In other words: I am not asking n8n to keep every successful execution. I am asking for the Wait / webhook-waiting resume mechanism to continue working when saveDataSuccessExecution: none, as it did in 2.26.9.