I use N8N in an ECS cluster and map have the following port mapping 80:5678.
The ECS cluster is part of a security group that allows all TCP traffic. Once I remove that and only allow 80, 443, 5678, 3306. N8N UI loads up correctly but gives me a connection timeout for the following endpoints :
Why is it that these endpoints require an inbound port that is not 80, 443, 5678 or 3306. What port is being used to call these endpoints?
Welcome to the community @yash!
There must be something else going on. n8n does definitely just use that one port and no other.
Thank you @jan
Is there an easy way to debug this? It’s weird that its only those 2 endpoints that are timed out. The rest load up fine, but the n8n UI fails to initialize correctly.
Hey @yash, welcome to our community forums.
Any chance you have the static (html, css and js) files cached using a CDN and / or a specific routes that redirects any
/rest/* calls to another group of servers?
I am trying to isolate the problem because as Jan said we only use port 5678.
Hey @krynble I don’t think it should have any LB rules or CDNs that route the calls to another group of servers. I am using docker so it should be a very straightforward implementation. I have allowed all traffic for the time being and it is working but will need to figure out why it was failing to load. I can try to reproduce it in an isolated env and provide more information on the same if it’s easier to debug that way.
If you hit F12 in your browser it would show you the port being used and what is being blocked although I suspect it could be something with how it is setup.
What environment options have you set up and does the ECS service have any options to set headers like a reverse proxy would do?