Removal of K3s based apps from TrueNAS

Related to the announcement posted @HoneyBadger, I find the direction taken by IX really bad. It does not makes any sense to start using Docker, we are going back to 2014. Kubernetes is here to stay and many companies are centralizing their work around it.

The worst part is the planning. IX implemented K3s, then after two major releases, we are told, hey, you can’t use K3s anymore but we let you use it as unsupported features. You can use Docker, which nobody on the planet implements into their software solutions anymore.

@Captain_Morgan we had in the past several discussions about K3s and it was exciting to see its adoption. Now we are brought back years in the past. Can we get some clarifications why this illogical decision was taken and more precisely, why it was done after two major releases, when everyone migrated their apps to K3s?

5 Likes

Four, actually. But iX thinks everyone wants Docker, and apparently the only reason they ever used k3s was for clustering, which they’ve given up on since the death of Gluster.

I think people should voice their concerns and clearly state that K3s should not become an unsupported feature. I find it simply ridiculous, nobody wants to use Docker. Show me a modern project based on Docker, today.

2 Likes

I tend to agree, but they seem to be in the habit of closing down threads that suggest it.

I don’t see why IX would close threads like this one. Is the customer that runs the show, not the developers. If they do that, they lack of transparency and prohibit to encourage the user base opinions, which is like a shot in your foot. People will just leave for other open-source solutions.

So they can look at other solutions, like Longhorn. Suse purchased the company so they can implement the solution everywhere in their tools. That would actually be a great implementation for K3s. That will solve all the mounting nightmares we have now in TrueNAS applications.

I know it’s not really adding to the conversation but “that’s bleeding edge for you”. However, can you not still use K3s even if it’s unsupported?

@John I would not use an unsupported feature, in a product that had it previously supported in production. Is really bad, from an end-user perspective. We are forced to use a product nobody on the planet uses anymore. I apologize, I don’t understand the bleeding edge context you mention. Using Kubernetes is a standard today, not Docker.

What are you using today?
How many systems… how many Apps?

What are your plans/requirements for the future?

‘bleeding edge’ as in the old saying; “they call it bleeding edge for a reason”. -i.e. if you want to use the latest “trend” you must expect to get cut sooner or later.

I am no Linux expert (I avoid it most times) but can you not just use podman and KD3?

We’re four major releases into the product’s lifecycle. How is that “bleeding edge”?

Kubernetes are still bleeding edge (it was a concept 9 years ago). Many new things adopt and want to consume them (chase after). Linux is bleeding edge because it tries to consume everything “new” it can; tons of developers told to ignore POSIX (a standard) and write something innovative first and foremost.

Kubernetes was released few months after Docker, September 2014. Hardly a bleeding edge technology, is a standard today. From your definition, Docker is also a bleeding edge product, even if nobody is using it today. By nobody, I talk about companies.

Morgan, I think is more from a technology perspective you have to look at it, not quantification. I understand that end-user needs always change during a product evolution. I might use 10 K3s applications today but decide to use 100, later on. Imagine the nightmare to wake up one day and be forced to migrate to a different solution, with no realistic alternative.

Right now, I’m forced to purchase additional hardware to run a Kubernetes cluster, which will fix permanently any possible issues that might arise from unsupported K3s in TrueNAS. I’m going to publish an open-source repository, for a proper K3s cluster setup, deployed with Ansible.

There is no way in the world I will use Docker, probably like 99% of TrueNAS Scale users. I have a difficult time understanding why would IX go backwards.

Correct. And wouldn’t a failed concept hurt a company if they adopted it and it disappeared one day just as much as an evolving concept could? How much have jails, and chroot changed in the last few (10’s of) years?

If k3/8s are not bleeding edge then you should be able to use it/them. How is the OS keeping you from using it?

My point is: K3s was adopted and announced with fanfare in Scale, devs were raving about it, as well the end-users. Two years or so later, we are told, hey forget about K3s, we are removing support and offer as alternative a solution that nobody uses anymore, you are in your own from now on.

It’s always difficult trying to judge prevalence from forum posts, but the “give us Docker” brigade have been very vocal here and on the old forum. What proportion of actual users do they represent? No idea.

By a similar token, my impression from the forums has been that most users have been using TrueCharts apps–but according to the telemetry data Kris mentions in the announcements, they represent a single-digit percentage of users.

1 Like

From my perspective, is not the number of users using a feature that matters in a product. Is the feature adoption itself, at global level. Is like saying, hey we are not upgrading anymore the OS in TrueNAS to new major versions because is pointless, nobody uses the underlying OS anyways, just the interface, we will do security patches and just move to major releases every 10 years. Is simply unrealistic.

Idea: use CORE?

1 Like

I moved to Scale to be free from old thinking jails approach and have Kubernetes as alternative. Looks like they plan to unify the jails with Docker and eventually get rid of Core.