Have recently deployed a new self hosted n8n. Has been getting stuck and frozen during workflows, the workflow runs but then it keeps running even when finished. Then I can’t save or do anything else without refreshing screen.
What is the error message (if any)? Have checked logs with no errors
Information on your n8n setup
n8n version: 1.63.4
Database: Postgres
n8n EXECUTIONS_PROCESS setting (default: own, main): I haven’t set this in any environment variables.
I’m sorry you’re getting some issues - thanks for bringing them up
I believe you mean that the executions keep running even though the UI displays the porgress as finished - is this it?
Does the last node produce an output as well?
You also mention that you can’t do much without refreshing - does refreshing trigger the executions to finish or do they still show as waiting after this?
So yes let’s say I’m testing a workflow, I proceed to the next step and it creates an output.
Then it just stays stuck on Executing workflow and I can’t do anything except refresh and start again from wherever it was pinned. You can see the symbol spinning in screenshot as if it is still waiting.
Also happens on different types of nodes for example GHL update contact or Airtable nodes doesn’t seem to care which node.
Happens randomly and on different workflows making it 10x more difficult to do anything. Any ideas please?
I have the same issue. Since I am completely new to n8n I don’t have any references to prior versions. Currently I am running self hosted on ver 1.63.4 via docker-compose. I am also using nginx reverse proxy. I had WSS errors too, but they could be resolved by adjusting the nginx.config file.
The status indicator gets stuck almost every time I test a node.
This sounds like an issue with the reverse proxy closing the websocket connection because of a configured timeout.
In your nginx config, in the relevant location block, please make sure all these lines are present:
Just as an update, Now when testing a workflow, often when the node finishes the progress just disappears and we’re back to wherever was pinned without touching anything.
In that case it sounds like Render is using a reverse proxy, and it might make sense to check their docs or their support for how to configure websocket timeouts to prevent the websocket connection from being closed.