n8n error handling comes down to three decisions for every workflow you publish. What a node should do when it fails, what the rest of the workflow should do after that, and who finds out when a run fails anyway. This guide walks through each one in the order you would set it up, using what n8n already provides, namely Retry On Fail, the On Error options, the Error Trigger, the Stop and Error node and the Executions list. It ends with how to test the setup and how to rerun a failed execution without doing the same work twice.
If you are new to the idea of triggers, nodes and executions, the guide on what n8n is used for and how a workflow runs explains them first. Everything below about n8n’s behaviour comes from the n8n documentation, read on 1 November 2026, with a link to each page.
What happens when a node fails
A failing node can stop the run. The docs page on working with nodes describes the On Error option Stop Workflow as one that “halts the entire workflow when an error occurs, preventing further node execution.” The execution is marked as failed and appears in the Executions list, as long as the workflow is set to save failed runs.
What does not happen on its own is that anyone hears about it. A workflow that fails at three in the morning stays failed until someone opens the Executions list. For a workflow that only you run, that may be acceptable. For one that takes orders, sends invoices or answers customers, it usually is not, which is why the rest of this guide exists.
It helps to sort failures into two kinds before choosing a tool. Some are temporary, such as a service that is briefly unavailable or a rate limit, and trying again a little later is the right answer. Others are permanent, such as an expired login, a missing field or a record that does not exist, and trying again will only fail again. Retries are for the first kind. Error outputs and error workflows are for the second.
Retry the step that failed
Nodes have a Retry On Fail option in their settings. According to the n8n page on handling API rate limits, “When you enable Retry On Fail, the node automatically tries the request again if it fails the first time.” You set how long to wait between tries in Wait Between Tries (ms). The same page says that if you are using retries to work around a rate limit, you should set the wait to more than the rate limit, for example 1000 for an API that allows one request per second. It also gives the message n8n shows when a service answers with error 429, which is “The service is receiving too many requests from you.”
Before you switch retries on, ask what the node does. Reading data is safe to repeat. Creating a record, charging a card or sending a message may not be, because a request can time out after the other system has already acted on it, and the retry then does it a second time. For nodes like that, prefer a step that first checks whether the record or message already exists, or leave retries off and let the error workflow alert a person.
Decide what a failure means for the rest of the run
The On Error setting on each node controls what happens after the node has failed. The nodes documentation lists three options.
- Stop Workflow halts the run. It is the safest choice when later steps depend on this one.
- Continue “proceeds to the next node despite the error, using the last valid data.” Later nodes then work with data that did not come from the failed step, which is easy to miss when you read the results.
- Continue (using error output) “continues workflow execution, passing error information to the next node for potential handling.” The node gets a separate error output, so you can connect it to an alert, a log sheet or a fallback path while items that succeeded carry on as normal.
The error output is usually the better of the two Continue options, because the failed items go somewhere you chose instead of disappearing into the normal path. A typical use is a loop over many records where one bad record should not stop the other ninety-nine, as long as that one record is logged for someone to fix.
There is a side effect to watch for. On the n8n community forum, a user whose error workflow never fired was told by another forum member that with Continue On Fail switched on the execution counts as successful, so the error workflow described in the next section is never called, and was advised to remove that setting from the nodes being tested (thread from January 2026). This is a forum report, not a statement in the n8n documentation, so check it on your own instance. If it holds for your setup, a node set to Continue needs its own alert in the workflow itself.

Send every failed run to an error workflow
An error workflow is a separate workflow that n8n runs whenever a workflow linked to it fails. The page on handling errors gracefully says it must start with the Error Trigger node and that one error workflow can serve several workflows. The setup has four steps.
- Create a new workflow and add the Error Trigger as its first node.
- Add the node that sends the alert, such as an email, Slack or Microsoft Teams node, and fill it with fields from the Error Trigger output.
- Name the workflow and save it. The Error Trigger documentation says a workflow that uses the Error Trigger does not have to be published.
- Open each workflow you want covered, go to its settings, choose the new workflow under Error workflow and save. The workflow settings page describes this option as “Select a workflow to trigger if the current workflow fails.”
The Error Trigger passes on details of the failed run. According to its documentation these include the execution ID and a link to the execution, the error message and stack trace, the last node that ran, the execution mode, and the workflow ID and name. A retryOf field appears only when the failed run was itself a retry. The execution ID and link are only there if the execution was saved.
One case looks different. If the failure happens in the trigger node of the main workflow, for example when the trigger cannot reach the app it watches, the data contains a trigger object with the error details and no execution information, because the workflow never started. Build the alert so that it still reads sensibly when the execution fields are empty.
What a useful alert contains
An alert is only useful if the person who receives it can act on it without opening five tabs. From the fields above, a short message with the workflow name, the last node that ran, the error message and the link to the execution covers most cases. The link matters most, because it opens the failed run with its data, which is where the fix starts.
Calculating End of Service for your employees?
Use our 100% free, private Gratuity Calculator updated for UAE Decree 33.
Calculate GratuitySend it to someone who owns the workflow and can fix it, not to a shared channel nobody reads. If a workflow handles customer data, think twice before copying input data into the alert itself. The link lets an authorised person see the details inside n8n without spreading them into email or chat.
Fail on purpose when the data is wrong
Some problems never raise an error. An order arrives without a quantity, a lookup returns no customer, or a total is zero. Every node succeeds, and the workflow quietly does the wrong thing. The Stop and Error node exists for this. Its documentation says it can “display custom error messages, cause executions to fail under certain conditions, and send custom error information to error workflows.” You choose between two error types, a plain Error Message or an Error Object written as JSON.
In practice you put an If node in front of it with the rule that must hold, and send the items that break it to Stop and Error with a message that says what was wrong, such as which field was missing. The run then fails, the error workflow fires, and the alert says exactly what to look at. The guide to n8n workflow examples for sales teams uses Stop and Error in the same way, so that a problem reaches a person instead of passing silently down the chain.
Test the error workflow before you rely on it
This step is easy to skip. In the forum thread above, the error workflow had been set up correctly all along, and the user had been testing in the editor. The Error Trigger documentation says plainly that “The Error Trigger only runs when an automatic workflow errors.” Running a workflow by hand in the editor will not call the error workflow, however badly it fails.
A simple way to test is a small throwaway workflow.
- Create a workflow with a Schedule Trigger set to run a few minutes from now, followed by a Stop and Error node with a message such as “test of the error workflow.”
- In its settings, choose your error workflow and save.
- Publish it so that the schedule runs it automatically.
- Wait for the run, check that the alert arrived with the workflow name, the error message and a link that opens the execution.
- Unpublish or delete the test workflow.
Repeat the test whenever you change the alert node or its credentials, because an error workflow whose alert node can no longer log in fails at the one moment you need it.
Fix and rerun a failed execution
Once an alert arrives, the Executions list is where you work. You can filter it by status and choose Failed, as described on the page for viewing executions for a single workflow. Two features help from there.
Debug in editor loads the failed run into the editor. The page on debugging executions says n8n copies the execution data into your current workflow and pins it in the first node, so you can fix the problem and run it again with the same data. It lists this as available on all n8n Cloud plans and on the Registered Community, Business and Enterprise self-hosted editions.
Retry reruns the failed execution. The page on viewing all executions describes two options. Retry with currently saved workflow uses your edited version, and Retry with original workflow reruns it unchanged. The same page explains several rules that decide whether a retry is safe.
- A retry resumes from the saved state rather than starting again from the trigger. Nodes that finished keep their results, and the failed node receives its saved input again and processes every item, including any it handled before it failed.
- If the failed node is an Execute Sub-workflow node, the whole sub-workflow runs again from its start.
- You can only retry an execution while n8n still stores it, so a workflow whose Save failed production executions setting is set to not save leaves nothing to retry.
- A succeeded execution cannot be retried, even if it did not do what you expected.
- Retry either the original execution or its failed retry, not both, because retrying both can repeat work on the same data.
The first rule is the one that causes duplicates. If a node sent messages to fifty customers and failed on the fifty-first, a retry sends the whole batch again. Before you retry, check what the failed node had already done, and if it creates records or sends messages, either remove the items that went through or make the node check first.
What these tools do not do
None of this fixes anything by itself. An error workflow tells someone that a run failed. Retries help with short outages and rate limits, not with wrong logins or bad data. Continue keeps a run going, and according to the forum report above it can also keep the error workflow from hearing about it. And the Executions list only holds what your settings tell it to keep. A setup that uses each tool for the failure it suits, and that has been tested with a real automatic run, leaves far fewer failures unnoticed than a workflow with none of it.
Where to go next
Webhooks are their own source of trouble, because a call that never starts a run cannot fail or send an alert. The Webhook node documentation says the test webhook stays active for 120 seconds after you select Listen for test event, and that a production webhook needs the workflow saved and published. A service still pointed at the test URL a day later is calling an address that no longer listens. How n8n webhooks behave, and the usual reasons they stop working, is the next guide in this branch of the n8n series.
If you build WhatsApp flows, the n8n WhatsApp bot tutorial using the official Cloud API explains why a bot that customers rely on needs an alert for failed runs. If you are moving workflows over from Zapier, how to hire an n8n developer and plan a Zapier migration treats error handling as a step of its own. And if you are still choosing a platform, n8n vs Zapier vs Make compares how each one counts the runs you are now protecting.
If you have workflows in production and want a second pair of eyes on how they handle failure, book a free 30-minute strategy call and bring a list of the workflows that matter most to your business.




