After upgrading n8n Docker (1.106 → 1.120.1), workflows reverted and ~2 weeks of changes disappeared (SQLite)

Questions
1

hi
I have a problem, my workflow loss after I upgeading my n8n version.
I’m self-hosting n8n using Docker Compose with SQLite (persisted in a Docker volume). After upgrading the n8n image from 1.106 → 1.120.1 and restarting the container, many workflows reverted to older versions, and some newer workflows disappeared entirely. This looks like a data rollback / “jump back in time” issue.

I’d really appreciate any guidance on possible root causes and recommended troubleshooting steps.

1) Environment / Setup

  • Deployment: Docker Compose, single n8n container

  • n8n version: 1.106 → 1.120.1

  • Image: n8nio/n8n:1.120.1

  • Database: SQLite (database.sqlite)

  • Persistent storage: Docker volume mounted to /home/node/.n8n

    • Volume: n8n_n8n_storage_test

    • DB path: /home/node/.n8n/database.sqlite

    • DB size: ~7.3GB

  • Verified mount: n8n_n8n_storage_test -> /home/node/.n8n (RW=true)

So this does not look like a missing/wrong volume mount.

2) Symptoms

After upgrade + restart:

  • Many workflows reverted to older versions

  • Some newer workflows are missing

  • The instance appears to have returned to an earlier state

3) Findings

I queried the persisted SQLite DB (workflow_entity) and found a missing time range in workflow updates.

updatedAt distribution:

  • 2025-12-31: 7 workflows

  • 2025-12-16: 52 workflows

  • No entries for 2025-12-17 ~ 2025-12-30

  • Earlier dates look normal

4) Questions

  1. Has anyone seen SQLite workflows/executions “jump back in time” after a n8n Docker image upgrade/restart?

  2. Given DB integrity is OK, what could cause a missing time range like this?

  3. What troubleshooting steps would you recommend to confirm the root cause?

Any insights would be greatly appreciated :folded_hands:
(Unfortunately, I don’t have backups for the missing workflows, so I’m focusing on understanding the cause and preventing it from happening again.)

2
  1. no, never
  2. I think it’s possible that Docker terminated abruptly and didn’t flush the data to disk; however, I have no idea if that would happen after so many days.
3

Probably not. Everything was fine before the upgrade. The rollback only happened right after the 1.106 → 1.120.1 restart, and the DB shows a ~2-week gap — way more than a typical “didn’t flush” scenario.

4

Hi @LiaoDDD,

Unfortunately I dont have answers to your specific issue, however I can provide some tips for future upgrades.

As a rule I make use of physical file system volumes instead of the Docker named volumes. This way it is easier to copy the n8n data folder as well as the database data folders when I make my backups before any upgrade. In case the above scenario happens, it is easier to restore and try again.

5

Thanks—right now I’m using the CLI to regularly export all workflows and credentials as a backup so that the same issue won’t happen again.

6

I would highly recommend you setting up daily server backups as the CLI exports only contain the workflow structures and NOT any links, credentials or execution history