Re: Docker

From: Paul Pathiakis <pathiaki2_at_yahoo.com>
Date: Thu, 13 Apr 2023 14:00:25 UTC
 I guess my opinion at this point is to drop this.  I don't see a valid point for diverting resources and various other things to accommodate 'docker' or many other things that are dependent on 'linuxisms'.  Where does it stop?  Do we start porting everything from Windows as well?  My point is there are many things in many OSes and variants thereof, that have hooks into proprietary parts of the kernel that are not 'modular'.  By modular, I mean that they can be compiled and used on another OS like most things in the ports/pkgs system.  Since this is 'kernel' level, I don't think FreeBSD should pursue such an endeavor with the limited resources at hand.  The FreeBSD kernel and userland are a thing of beauty and refinement imho.  All I have to do is look at the CVE database to see that in the last 10 years there only a couple of hundred bugs.  Just the linux KERNEL has 1000s as does windows.  I would worry that anything that had ties into the Linux kernel is probably an issue waiting to happen.
I've been doing system administration and system architecture for over 35 years...  When people ask what the dominant *nix OS is and are expecting Linux.... It starts us down the road of all the big boys use FreeBSD because they can't afford to have constant patching and vulnerabilities.
So, it's either in a hypervisor and we go from there or drop it.  The amount of time spent on this discussion is becoming 'trollish'
Paul

    On Thursday, April 13, 2023 at 08:23:35 AM GMT-5, Mario Marietto <marietto2008@gmail.com> wrote:  
 
 ---> Couldn't we just run docker on bhyve?
more no than yes. You could try to put yourself in other people's shoes. You are only moving the problem. You are indirectly asking the users that come from another system to learn bhyve if they want to use docker. Why should they learn something different to just use what they need ? At this point they could jump directly to learn jails,instead of bhyve and / or docker. To learn something different requires time,energy,etc. This is not a good business card for the new users. And it implicitly admits that a useful and popular tool like docker doesn't work on an efficient operating system like FreeBSD. Yes there are great tools like docker for freebsd, but those users don't need it, they just want docker. Maybe they don't even need to learn bhyve. Just Docker. Your reasoning is typical of someone who has been using freebsd for some time, you don't think like those users who would like to adopt it and are evaluating the pros and cons. Take also in consideration that running bhyve to run Docker is a waste of resources on the machine,if I want to run only Docker,because in a normal situation,I shouldn't have the need to use bhyve. Users that have already boarded FreeBSD have probably already come to appreciate jails and many of them don't need to run bhyve to get docker. Remember the focus of my argumentation : it is something like this : I offer a native implementation of docker on FreeBSD and I use it as bait to attract more users. And between those users maybe there will be also good developers that will love FreeBSD even for different reasons than docker. The ultimate goal is to make freebsd a little more attractive to the industry, because as far as I read, it's slowly disappearing.


On Thu, Apr 13, 2023 at 2:59 PM Miguel C <miguelmclara@gmail.com> wrote:

100% Agree with this, and the fact is there have been cases where there is that tolerance and there are maintainers making efforts to bring "linux" things to freeBSD even if via linux emulation.
Docker has been mentioned many times in mailing lists and forums and there is always comments like "but why jails are much better" etc, sometimes not only intolerant but rude reply that serve only to drive people away IMHO.
I also don't get why is that so complicated, is it just cause FreeBSD's maintainers/community don't want to even consider docker on FreeBSD? Couldn't we just run docker on bhyve? I'm sure it would serve the "just want to test this image purpose" but I suspect there will be some issues with Filesytem/network, not issues per say, but more like it likely takes some work to get this to run in easy manner, but I think I've seen mentions of using sshfs or zvols to make this part easier.
MacOS and Windows use virtualization anyway, sure Docker "DESKTOP" is supported but docker, but they are still using a VM at the end of the day and handle the filesystem/network stuff for the user.

I've never tried this my self but I don't think it should be that super complicated unless you plan to run docker on prod envs, I think here, the argument that "right tool for the job" is very valid.... I use docker on my macOS but I'm not going to run things in prod in macbooks ofc, I will still use Linux, K8s etc.

Perhaps the FreeBSD foundation could invest a bit in getting a tool to easy the way of running docker through bhyve, I do believe this would be good for user adoption, but probably there are other priorities.



On Thu, Apr 13, 2023 at 12:32 PM Mario Marietto <marietto2008@gmail.com> wrote:

The point of my argumentation is not if FreeBSD has or not good tools for containerizing and securing applications. It has. Point is that the users that don't know FreeBSD are tied to their own tools and rarely want to change them. Almost everyone wants to change. But trying,experimenting and changing something in the workflow is important,because every tool has bad and good sides. There are many docker images already to be used on the net and this will save a lot of time and effort and money for a lot of people. This is a fact. And I think that it happened because Docker is...good. FreeBSD has tools like docker,but the mass production of containerized images never happened. So,would we ask ourselves the reason ? Maybe something has not gone well. I use Linux and FreeBSD and I "love" both these systems. Linux has a larger user base than FreeBSD. A larger user base may mean more innovations in a small time,a faster bug correction and so on. 

I think that mostly advantages from the implementation of docker on FreeBSD will come from the user base. Mostly for those users that come from linux or other OS and that already use docker and kubernetes. I don't think those users are a small number. Those users could jump to FreeBSD if Docker / Kubernetes are implemented in FreeBSD. This could be the straw that broke the camel's back. You argue that the jails are working already great and that they should use them. I argue that the freebsd community could have a more tolerant behavior to the users that could jump to the FreeBSD world and they should not force them to learn only new technologies at first. To have some important tools which work on multiple systems means having a good business card. So,in the end I ask to myself and to you : FreeBSD needs to grow in terms of community ? Does it need to be populated by a bigger number of users that will come from another OS base community ? 
On Thu, Apr 13, 2023 at 10:17 AM Alejandro Imass <aimass@yabarana.com> wrote:



On Wed, Apr 12, 2023 at 4:28 PM Paul Pathiakis <pathiaki2@yahoo.com> wrote:

 I believe the simplest thing would be to wrap jails or iocage in an interface that looks like and behaves Docker-like.


and Bastille! 





-- 
Mario.




-- 
Mario.