Is selfhosted n8n more bugged than paid version?

[funkvay]

Describe the problem/error/question

I’ve been using the self-hosted version of n8n and noticed that when creating workflows, especially with new nodes, I frequently encounter bugs that I can’t resolve. I want to understand if these issues are specific to self-hosted instances or if they also occur in the paid, hosted version.

If switching to the paid version would provide a more stable experience, I’d consider it. But if the same bugs exist across both versions, would it be better to wait for fixes from the developers?

I’m just trying to find the best option for working efficiently with n8n. Could the problem be related to the OS where I host? I want to test a more stable setup but not through the 14-day trial, as I only have time to test once a week or once every two weeks - this is why I’m self-hosting.

Information on your n8n setup

• n8n version: 1.80.3
• Database: SQLite (default)
• n8n EXECUTIONS_PROCESS setting: main (default)
• Running n8n via: Docker Compose
• Operating system: Ubuntu 22.04 with Docker

[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:

[BramKn]

Not perse, but of course if you host yourself you need to make sure your own server is stable. For cloud(assuming you mean that by paid) it is done for you. And of course the n8n team knows how to make their own platform stable.
So no there should not really br a difference.

What bugs are you talking about though?
Also recommend using postgres for your n8n database from the start. Sqlite is nice and all but really not worth the hassle when you need to switch over to postgres. (you need postgres for heavier worklpads and scaling) I dont use sqlite, so don’t actually know if this could cause stability issues.

[funkvay]

I left a comment on the open GitHub issue Can’t use Window Buffer Memory - undefined session id #13295. It’s not my issue, but I added details since it seemed related. The developers are already reviewing it.

I also ran into an error with the AI Agent node “No prompt specified” (attached screenshot)

, but I found a workaround on the forum. Additionally, I had a Telegram trigger issue “bad webhook: an https URL must be provided”, which I resolved using ngrok. Could ngrok have caused any issues? I know this is a broad question, but I’m asking in case one of the developers has encountered something similar.

Since these issues either have workarounds or are already being worked on by the developers, my question was more about overall stability. The project is really cool and has a lot of potential, and I’d love to dive deeper into it. I just want to make sure I’m not spending hours troubleshooting issues that others might resolve in minutes.

Also, I initially ran n8n using Docker Compose, not just docker run, so I’ll test running it as a regular container to see if that makes a difference. Maybe that setup plays a role too.

Thanks for the PostgreSQL advice! I’ll make the switch and see if n8n runs more smoothly. I’ll leave a comment here after testing everything - maybe it will be helpful for others.

[BramKn]

Ok, so my first answer was still right. No there is no issues on stability selfhosting vs cloud, but yes it can be caused by how you set it up when selfhosting.

Seems like some of your issues are due to the fact you are not setting n8n up right. When setting up a production environment you should not use ngrok. Or atleast not the free variant where your URL changes everytime you restart as far as I know.

Your screenshot however has nothing to do with your setup, but more with your workflow design. You can already see the expression you put in is undefined (the field you are listing). So this will of course result in no prompt given.
Have you tried doing the courses first?

It seems that except the one issue you mentioned the github issue for all other issues are user error, so not really bugs.
I am not trying to tell you to “get gut”, it is just something you need to realise.
Hosting things can be tricky and requires experience to do it properly without issues. Therefor Cloud is an easier option, and more stable if you do not have the experience.
Most people find n8n is a bit more tricky to get into compared to Zapier and the like, I had the same thing when I started.
A lot of people just dive straight in which results in a less than smooth experience, because they simply do not understand first before trying all kinds of things. And then we get a lot of questions and some people blaming the product(not saying you do that perse), while it would have all been solved if they just did the courses first.

[funkvay]

Thanks for the detailed response! I get your point about self-hosting - I’ll take a closer look at that. Also, I didn’t realize that ngrok might not be ideal, so I’ll reconsider that approach.

As for the AI Agent node error, I asked because that’s the default value I get every time I create the node. When I look at videos and tutorials, others don’t seem to have that issue, so I wasn’t sure if it was something on my end or a common problem. I even used their json files to make sure that I’m doing everything right, but still got that error.

That said, I’m actually glad to hear that the issue isn’t with n8n itself but rather my hosting setup - means I can dig deeper and improve it. So thanks again for the clarification! You’re right, I’ve worked with database integrations, background services, and system-level stuff, but setting up web services like this is new for me. Never used ngrok before either - just found a quick fix and went with it.

As for the webhook issue without needing a Public HTTPS Domain, I actually came across this thread (no idea how I missed it for a week). Just leaving it here in case it helps others too. But I’m still curious - do you have any other solutions aside from what’s mentioned there? Just asking out of interest, no worries, I’ll handle it myself. Just wondering how someone more experienced would approach it - of course, if you use the Telegram bot functionality at all.

Ngrok seemed solid at first, but from what you said, it does sound like it could cause issues, even if not directly related to what I’ve run into. I guess the best option I have right now is reverse proxy with Nginx to manage SSL termination. Although it’s likely to be quite difficult if the commenter is to be believed, it’s still worth a try.

Overall, this product is awesome, and the fact that it’s open source makes it even better. I really appreciate the work the developers and contributors put into it - huge respect to the whole team.

And thanks again for your answer!