Node connections disappear randomly self-hosted

[castigation]

We have an issue with workflows losing connections between nodes, usually at random times between executions. Even on workflows that haven’t been updated in some time. Still can’t figure out what the issue is, seemingly no way to replicate this.

Only thing that I can share is this

workflow 1
before:

after:

workflow 2
before:

after:

Still the same theme, that randomly the connections between nodes disappear.

"connections": {},

debug info:

[details="instance information"]
# Debug info

## core

- n8nVersion: 1.72.1
- platform: docker (self-hosted)
- nodeJsVersion: 20.18.0
- database: sqlite
- executionMode: regular
- concurrency: -1
- license: enterprise (production)
- consumerId: 584a6d02-5740-425e-ad6a-512ed2b32f08

## storage

- success: all
- error: all
- progress: false
- manual: true
- binaryMode: memory

## pruning

- enabled: true
- maxAge: 336 hours
- maxCount: 10000 executions

## client

- userAgent: mozilla/5.0 (macintosh; intel mac os x 10.15; rv:133.0) gecko/20100101 firefox/133.0
- isTouchDevice: false

Generated at: 2025-01-04T16:37:23.318Z
[/details]

• **n8n version:**1.72.1
• Database (default: SQLite): default
• n8n EXECUTIONS_PROCESS setting (default: own, main): default
• Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
• Operating system: Ubuntu Server Hetzner

Please help someone. Currently the only fix for us has been to copy+paste a previous executions workflow into the editor and then save. But this happens again after some time so it is very scary.

[n8n]

It looks like your topic is missing some important information. Could you provide the following if applicable.

• n8n version:
• Database (default: SQLite):
• n8n EXECUTIONS_PROCESS setting (default: own, main):
• Running n8n via (Docker, npm, n8n cloud, desktop app):
• Operating system:

[castigation]

Update: We think this has got to do something with the “Execute Workflow” node which is the common denominator with the errors.

[gualter]

hi @castigation

We have had a few reported issues similar to this in the past, where learning the cache may have helped. Can you exclude that by trying a new or private session on that browser or a different one?

Also, is it possible to update to the latest version of n8n just in case?

Thanks!

[castigation]

Hi gualter,

Yes we did try many different browsers on Windows and macOS but we think this has more to do with the Execute Workflow node, especially when you Execute another workflow outside of the current workflow. In that case when you have many concurrent workflows running, somehow the connections get lost somewhere and this causes all the next executions resulting in empty.

The fix for us has been to replace all the Execute Workflow nodes with HTTP Requests and Webhooks. So when we want to call another workflow, we are calling a webhook instead. Also the Respond to webhook node is useful, because then it will error out in case the connections get lost again - which hasn’t happened yet.

Thanks for reaching out, we appreciate it since this is a very rare but very irritating bug…

[gualter]

Thanks for the walkthrough. Without a proper way to replicate this, it’s going to be hard to fix it

Does the current implementation you have work for you?

[ivov]

@castigation Would you be able to put together a minimal repro for this? At least one where you suspect this is likely to happen, even if not consistently.

One idea would be to set these env vars:

export DB_LOGGING_ENABLED=true
export DB_LOGGING_OPTIONS=query
export N8N_LOG_LEVEL=debug

Assuming your theory is right, once this disconnection happens, we can use the execution ID and workflow ID and the timestamp to locate the relevant app and DB logs, to pinpoint where the connections are persisted as an empty object.