When an n8n workflow fails, the executions list shows which one and at what node. From there you have three tools to make it robust: retrying on the node, deciding what to do on error, and an error workflow that sends you a notice. n8n is third-party software running on your VPS; the exact screen names depend on the version, so confirm in the project’s documentation.
First, read the error
| 1 |
Open the executions list and find the one marked as failed. Open it.
|
|
| 2 |
Find the red node. It is the one that failed. The nodes before it ran fine.
|
|
| 3 |
Read the message and the data that went into that node. It nearly always explains: an empty field, a refused credential, a request that took too long.
|
|
| 4 |
Reproduce it with the same data. You can run just that node, to test the fix without repeating the whole workflow.
|
|
The three defences
| Defence |
What it is for |
When to use it |
| Retry on fail (in the node’s settings) |
Tries the same node again a few times, with a pause between tries. |
Passing failures: a slow network, a busy API, a request limit. |
| What to do on error (in the node’s settings) |
Stop the workflow, or carry on with the rest of the data and handle the error separately. |
When an error on one row should not stop the others. |
| Error workflow |
A separate workflow, with the error trigger, that runs when another one fails. |
To get the notice: e-mail, message, a row in a sheet. |
Building the alert
| 1 |
Create a new workflow that starts with the error trigger. Then add a node that sends you a message (e-mail or Telegram). See alerts to Telegram.
|
|
| 2 |
Put in the notice the name of the workflow that failed, the node, the error message and the time.
|
|
| 3 |
In the settings of each important workflow, choose that workflow as the error workflow. Link it to all of them, or the ones you miss fail silently.
|
|
| 4 |
Cause a failure on purpose in a test workflow and see whether the notice arrives.
|
|
|
Retrying is not for everything. If the node sends a message or creates an order, retrying can duplicate the effect: the customer gets it twice. On those nodes, retry only if the service accepts an identifier that prevents duplicates, or handle the error by hand.
|
If n8n stops answering or restarts by itself, the problem may not be the workflow: see n8n running out of memory: the causes. If the trigger never fires, see n8n webhooks: test URL, production URL, and why they do not fire.
|
Clear old executions. They keep the data that passed through, and that fills the disk and holds customer data longer than you need. Set how long you keep them.
|
RECOMMENDED PRODUCT Web hosting with cPanel Domain and SSL included, daily backups and the panel you already know. from $6.60/mo (3-year plan, with coupon) See plans |
Angola
Brasil
Kenya
Moçambique
Nigeria
Portugal
South Africa
United Kingdom
United States
International
