Hacker News Reader

Comment by e145bc455f1 | Hacker News Reader

Are there any use cases where i should prefer FreeBSD over Linux?

I have been using simultaneously both FreeBSD and Linux for more than two decades.

I use Linux on all my laptops and desktops and also on servers that are used for computational purposes, i.e. on which I compile software projects or I run simulations or FPGA syntheses or CUDA programs on GPUs.

I use FreeBSD on servers that implement various networking or storage functions, e.g. firewalls, NAT, DHCP servers, routers, switches, file servers, backup on magnetic tapes, DNS servers and proxies, e-mail servers, Web servers and proxies and so on.

The main reason why I prefer FreeBSD in such servers is that for it I need to waste less time for maintenance and monitoring, because it provides a more stable and consistent environment where I need to make very few custom changes for the services that I need.

Why not Debian there? With unattended upgrades setup you can install and forget them.

In FreeBSD many services that I need work immediately after installation with very few changes in the default configuration.

In Linux I always have to work much more for a sane configuration.

There are also various specific details, e.g. I like much more the firewall programs that are available for FreeBSD than those that are available for Linux, so for any computer facing the public Internet I use FreeBSD. FreeBSD has better support for SCSI/SAS devices, so where I use such devices, e.g. tape drives, FreeBSD is also the only choice for me.

In FreeBSD I also normally use unattended upgrades and I almost never use more than an hour per year for maintenance, except for reading my e-mails where I receive statistics from the logs of the servers.

If you want to a release a new proprietary hardware and you need a full fledged general purpose OS without releasing the source code. See Sony > PlayStation 3/4/5

ZFS, which is natively supported on FreeBSD. Jails, maybe; they are nice to have, certainly.

Beyond that - for regular server use cases, at least - it's mostly a matter of taste, I think. There are situations where you need an enterprise distro, i.e. RHEL or SLES because some third-party software requires that or regulatory compliance requirements demand it. Most of the time, I think the difference is minor for most intents and purposes.

EDIT: If you strongly dislike systemd, FreeBSD is an attractive choice, but there are Linux distros that use a different init system, too.

What are the FreeBSD/OpenBSD init systems? Are there competing ones?

I like runit, which is actually from Void Linux but can be used on FreeBSD. There's also OpenRC, again not limited to BSDs.

The BSD systems have their own init system, called rc. I've never written an rc script, but as a user/admin, I find it very nice.

There was an effort to create something akin to macOS's launchd for FreeBSD, but I don't know how that worked out.

> I've never written an rc script

Not trying to "RTFM" you but https://man.freebsd.org/cgi/man.cgi?query=rc contains a nice minimal example script for anyone who's curious.

Thank you! I recently wrote my first systemd unit file, and the experience was mostly pain free. I might recreate it as an rc script for good measure.

For the more advanced stuff don't miss the link at the bottom that's unfortunately broken in this web manpage reader https://docs.freebsd.org/en/articles/rc-scripting/

Here's a bit on OpenBSD :

https://www.bsdfrog.org/pub/events/openbsd-rcd-EuroBSDcon201...

In my experience, if you know what you're looking for, and FreeBSD meets your needs, and you prefer a more classic unix-y setup, then it's great. If you want the cutting edge of everything, it may still work for you, but it's worth investigating to be sure. FreeBSD doesn't always do bleeding-edge stuff without a bit of hand-holding.

With ZFS and boot environments, it's great for upgrades, if you're okay with some downtime. You can certainly do high-availability, but again, there's some hand-holding.

I much prefer FreeBSD for everything I can use it for. But I don't use Docker, or Kubernetes. I use it as a server OS where the only interface is ssh.

It works well as a general purpose desktop OS as well as rock solid server. It is arguably simple and less complex, and has a cleaner, compact and immensely readable source code as compared to Linux.

Serving static web content via kTLS. We (at Netflix) have worked for a decade to make this as efficient as possible.

“Prefer” is a subjective term. For the same use case some people might prefer FreeBSD, while others will prefer Linux.

Personally I prefer FreeBSD as an operating system but the lack of Docker keeps me stuck to Linux (I don’t even like Docker but at this stage it’s now the de facto standard for packaging containers).

True, but one of the advantages of the BSDs is that they provide an alternative way of doing things. For FreeBSD, you have Jails with Pot and Potluck and other alternatives. It might not be the de facto way of doing things, but the BSDs have never really been keen on following the standard way at the cost of everything else.

I’ve used jails since FeeeBSD 7 or 8 and love them. For me, the power of Docker is less about the OCI format (whatever it’s called) and more about Dockerfiles. If Jails could work against Dockerfiles then that would be a massive step forward.

Though to be fair I haven’t tried Pot. Used ez_jails a lot in the past, as well as rolling my own build scripts. So maybe I’m missing a trick?

This is what has been scaring me about Linux in the last 10 years as it seems we have been talking mostly about containers & ended up in a strange place:

- a lot of project distribute a docker image or we have community "distributions" like linuxserver.io. But nothing come close to a good distribution package management so far,

- the security of containers is still very unclear, misunderstood & unequal,

- the ecosystem is now fragmented between docker, podman, etc.

- generally it feels like things have become more complicated !

We also now do see new generation (?) of distributions centered around containers like CoreOS, MicroOS or even Silverblue.

My questions would be: is the BSD world trying go move this way too with bsdpot & potluck or consider this is not the way to go or this just a distraction and Ports will always be the way to distribute software?

I run FreeBSD as my host and pass data into linux VMs with bhyve 9p protocol support. This way I can have my docker data on my FreeBSD host's ZFS. Just pointing out another option, but it also depends on what type of stuff you are doing.

This sounds interesting. Got any more details about it?

I'll give a few pointers, I have done this with Debian and NixOS as well as docker and podman, currently I use NixOS/podman and that is all I have for examples. If you were using debian you would use a mount command with the Fstype and options below.

In a bhyve vm config file you add 'bhyve_options' and add some virtio-9p mount points. they are 'per slot' so each of these mounts will need a new number '-s 15' '-s 16' as in my below example.

on FreeBSD host, assuming you are using bhyve-vm (works with plain bhyve also):

/usr/local/vm/nixos/nixos.conf

bhyve_options="-s 15,virtio-9p,podman-archivebox=/HOSTZFS/Services/VM/NixOS/ArchiveBox -s 16,virtio-9p,podman-music=/HOSTZFS/Audio/Music/Flac"

in VM (nixos):

  fileSystems."/mnt/podman-archivebox" = {
      device = "podman-archivebox";
      fsType = "9p";
      options = [ "trans=virtio" "version=9p2000.L" "_netdev" "cache=loose" ];
  };

  fileSystems."/mnt/podman-music" = {
      device = "podman-music";
      fsType = "9p";
      options = [ "trans=virtio" "version=9p2000.L" "_netdev" "cache=loose" "ro" ];
  };

If you were not using NixOS, add your mount points in the VM (/mnt/podman-archivebox, /mnt/podman-music). Then mount them with your docker/podman setup.

  virtualisation.oci-containers.containers."ArchiveBox" = {
    image = "archivebox/archivebox";
    ...
    ...
    volumes = [
      "archivebox_data:/data"
      "/mnt/podman-archivebox/archive:/data/archive"
      "/mnt/podman-archivebox/sources:/data/sources"
      "/mnt/podman-archivebox/ArchiveBox.conf:/data/ArchiveBox.conf"
    ];

  virtualisation.oci-containers.containers."navidrome" = {
    image = "deluan/navidrome:latest";
    ...
    ...
    volumes = [
      "navidrome_data:/data"
      "/mnt/podman-music:/music:ro"
    ];
  };

The 9p mount points won't get passed into the VM until you STOP the VM and then START the VM (restart won't work). Once that is done and you can confirm the files are passed through and working, things 'just work'. This does not work for things that require file locks like passing database files through 9p. Hope this is helpful to someone.

Thanks for sharing :)

It's the other way 'round: You should use FreeBSD unless and until you have a use case that requires Linux. (For production; for yourself do what you like.)

If you want something with proper linage from UNIX, that was part of the whole UNIX System V vs BSD wars drama, the related AT&T lawsuit after they were allowed to charge for UNIX, 386BSD being reborn.

Or just a kernel that happens to have been inspired by UNIX and took its own path afterwards.

I’d say stuff that you just want to work, maybe because you’re a startup or otherwise small shop - in contrast to stuff where you want to generate a lot of artificial work for devops and consultants, for example in a corporate environment with K8s ;)

Something something ZFS

I've met people running ZFS on Linux quite reliably this days, bar Oracle.

ZFS is nicer on FreeBSD. Even on the more ZFS-friendly distros it still feels like an after thought. But in FreeBSD it’s as close to being a core part of the OS as it has been since the demise of Solaris and its various OpenSolaris forks.

I say this as someone running ZFS on Linux too. So seen things from both sides of the fence.

It’s hard to explain why it’s a nicer experience though. Maybe someone else can?

Then again something something inotify, which kqueue does not equal.

I'm a bsd lover so don't take that wrong.

One example of the nicer experience, maybe as a by-product of the 1st class ZFS support, is that updating FreeBSD doesn't make me nervous like updating the Linux kernel does. On Linux, you're mostly limited to the kernel versions that have been explicitly tested against ZFS.

ZFS memory stats are integrated into FreeBSD's top. Boot environments are trivial to configure; the OS upgrade tools integrate with BEs for you. It is considered a standard configuration (and has been so for years), so guides, docs, and software are written with explicit integrated support and expectations that you'll use it.

Reply on news.ycombinator.com