CRM and automation

n8n 3.0: what to check before upgrading your workflows

By Edgar Jaimes 4 min read

Laptop on a wooden desk with code open on the screen

n8n 3.0 brings breaking changes: removed nodes, mandatory Docker and new variables. Review this checklist before upgrading your automation.

You have a workflow that has been passing form leads to your CRM for months without failing. One Friday someone upgrades n8n "to get the latest," and on Monday the workflow won't start because it used a node that no longer exists. With a major version like 3.0, that risk is real.

This guide summarizes what the official documentation asks you to review, so you can upgrade calmly and without surprises.

What is n8n 3.0 and why does it matter?

It is a major n8n release that removes old nodes, changes default settings and requires a different deployment for anyone who hosts it themselves. The version 3.0 changes page states that, for now, it is an early access release and gives no release date. That is why it pays to prepare now and not rush to upgrade.

If you don't use the tool yet, start with what n8n is and how it automates without code.

Laptop with code on the screen next to a plant on a desk
Before touching production, check which parts of your workflows depend on old pieces. Photo: Unsplash

How do I know which workflows will break?

Open Settings > Migration Report on your instance. According to the n8n documentation, that report lists the workflows and settings that need attention before upgrading, and it can also migrate some workflows, such as those using the old AI Transform and Edit Image nodes.

Do it as a first step and save the result: it is your to-do list.

Common mistake: upgrading the production instance directly. Test first on a copy, with your workflows imported, and confirm that the most important ones (leads, payments, alerts) keep running.

Which nodes and features disappear?

Several old nodes disappear, and workflows that use them fail when they run. The documentation clarifies that published workflows with a removed node don't even activate at startup. These are some of the most relevant changes:

What changesWhat you should do
Function and Function Item, Cron and Interval, HTML Extract, Read PDF nodes and othersReplace each one with the replacement the documentation indicates
AI Agent version 1 and its old modesUse Tools Agent in its latest version
Old OpenAI nodesReplace them with the current nodes
Execute Sub-workflow with the Local File or URL sourceUse the Database or Define Below source
Edit Image versions 1 and 1.1Move to version 2, which may change the output

If your automation uses AI agents, also check how to write good instructions for your agent, because when you migrate versions it pays to test them again.

What changes if I self-host n8n?

You have to use a Docker-based deployment: the documentation states that installations with npm and npx n8n are no longer supported. There are also configuration changes worth noting:

  • The ~/.n8n/binaryData folder is renamed ~/.n8n/storage, and n8n won't start if both exist.
  • N8N_RUNNERS_TASK_TIMEOUT drops from 300 to 60 seconds.
  • Encryption key rotation is enabled by default.
  • The Slack API credential requires a signing secret, and a workflow with a Slack Trigger won't publish until you add it.
Blue network cables connected to a numbered patch panel
The server where your automation lives changes too: in 3.0 it moves to Docker. Photo: Unsplash

Which behaviors change without warning?

Some workflows don't fail, but they behave differently. They are the most dangerous because nobody sees an error. The documentation mentions, for example, that the Gmail Trigger node skips drafts unless you turn on Include Drafts, and that sent or scheduled emails no longer trigger the workflow. How Always Output Data responds on nodes with several outputs, such as If and Switch, also changes.

That is why it pays to have alerts when an n8n workflow fails before migrating, not after.

How do I upgrade safely?

Follow this order: report, copy, test and only then production.

  1. Open Settings > Migration Report and note what is pending.
  2. Make a copy of your instance and your credentials.
  3. Fix or replace the removed nodes in the copy.
  4. Run the critical workflows with test data.
  5. Upgrade production during low-traffic hours, with someone on standby.

If your workflows depend on calls between systems, also review what webhooks are and how to use them with n8n.

In short

n8n 3.0 is not an upgrade to do on a Friday afternoon. It is like the Monday workflow that wouldn't start: what works today may depend on a piece that no longer exists. Review the migration report, test on a copy and upgrade with a plan. If you would rather have someone do it for you, check out our n8n automation service.

Frequently asked questions

What is n8n 3.0?

It is a major n8n release with breaking changes. The official documentation describes it as an early access release and lists the changes you should review before upgrading.

How do I know if my workflows will be affected?

Open Settings > Migration Report. According to the n8n documentation, that report lists the workflows and settings on your instance that need attention, and it can migrate some workflows, such as those using AI Transform and Edit Image.

What happens to npm installations?

The documentation states that self-hosted n8n must use a Docker-based deployment. Installations with npm and npx n8n are no longer supported.

What happens to a workflow that uses a removed node?

According to the documentation, workflows with removed nodes fail when they run, and published ones stop activating when the instance starts.

Should I upgrade right away?

No. First review the migration report, test on a copy of your instance, and only then upgrade production.

Do not stop at the theory: apply it to your business

We implement n8n automation end to end, with your real operation. The first conversation is free.

Keep reading