n8n 3.0 beta is out and the upgrade is one way. Run the Migration Report on every client instance before December

n8n 3.0 is a cleanup release, and cleanup releases are the ones that quietly break client workflows.

What happened

n8n released the 3.0.0 beta on October 27, with the stable release planned for early December. The company says it held its breaking changes for a major version, so 3.0 focuses on reliability, stability and security rather than new features.

The main changes:

• Self-hosted n8n now requires Docker. Installing with npm or npx is no longer supported. Kubernetes and Helm setups are not affected.
• Legacy nodes are removed, including Function, Function Item, Item Lists, Cron, Interval, the legacy OpenAI node and version 1 of the AI Agent node with its older modes. Workflows that use a removed node fail.
• The default Code node timeout drops from 5 minutes to 1 minute.
• Slack credentials now need a signing secret, and unverified community packages are off by default.
• Gmail Trigger versions below 1.4 now run with 1.4 behavior: a default of 10 emails per poll, drafts skipped unless you turn them on, and sent emails outside the inbox no longer trigger.

Instances on 2.39 or later get a Migration Report under Settings. It lists affected workflows, fixes many automatically as new workflow versions to review, and keeps published workflows on v2 while you work. The upgrade is one way, with no rollback except restoring a backup. After stable ships, 2.x gets security and critical fixes for 12 months, and 1.x support ends December 31, 2026.

My take

Treat this like a production migration, not a version bump. The risky changes are not the removed nodes, because the report flags those. The risky ones are behavior changes that still run but give different results.

The Gmail Trigger change is the one I would check first. An intake workflow that relied on catching drafts or sent mail will simply stop seeing them, with no error. The 1 minute Code node timeout will hit long data cleanup or batch scripts. And the Always Output Data change on If and Switch nodes can change which branches run.

My order of work: update to the latest 2.x, run the report, fix or mark won't fix, back up, then upgrade a staging copy and run real sample data through every client workflow before touching production. If a client still runs n8n with npx on a small VPS, plan the move to Docker first, since that is a hosting change, not a workflow change.