Handle Errors, Duplicates, and Approval

Today's Goals

1. Separate retryable failures from human-handled failures.
2. Prevent duplicate sends across repeated executions.
3. Review execution records and rollback steps before activation.

A workflow can be correct on its happy path yet fail in production because an API times out, a trigger repeats an event, or an operator reruns a failed execution. Reliability is the set of rules for those ordinary conditions. Today's output is a reviewed operating plan for the fictional triage flow.

Plain-Language Walkthrough

Give failures a destination

The n8n error-handling guide explains ways to handle node errors and error workflows. Decide which errors may retry, which should pause, and which should enter manual review. A temporary read-only lookup failure may retry a bounded number of times. An authentication failure needs a credential owner. A parser error or unsupported quote sends the ticket to a human with the raw source ID and failure reason. Never treat “continue on fail” as permission to pass an empty result to the next action node.

The executions view is where an operator can inspect runs and, when supported, retry a failed one. Record the execution ID, workflow version, source ticket ID, failed node, and whether any side effect occurred. Keep only the minimum customer data needed for diagnosis. The operator should be able to distinguish “no draft created” from “draft created but confirmation failed.” Those states require different recovery steps.

Sign in and link WeChat to read

Sign in, then link your WeChat mini program account on the account page. Linked accounts share one free article per day. Unlock additional articles with rewarded videos in the mini program.

Scan in WeChat to open Dayward AI. Open the same course and chapter, choose the rewarded video, then refresh this page after completion.

Unlocked — refresh to readSign in / scan with WeChat →