> The largest failure was with btrfs — after a reboot, a 50 TB filesystem (in mirror, for backups) simply stopped working.
RAID is not a backup.
31–40 of 91 posts
> The largest failure was with btrfs — after a reboot, a 50 TB filesystem (in mirror, for backups) simply stopped working.
RAID is not a backup.
It’s kind of a ridiculous thing to claim without very substantial proof.
I think it’s made up to justify installing BSD.
I don’t buy this article. > Clients are often influenced by hype. A few years ago, it was "Linux is a toy." Now, it's "Why bhyve and not Proxmox?" They ask, "How can they sell FreeBSD? There's no AI, there's no Cloud, there's no Kubernetes, there's no blockchain – there's nothing!" I am very confident that this is more ‘fan fiction’ than the author would like to admit. The sort of hypothetical that someone cooks up o…
I made a semi-successful blockchain based product 5 years ago. At two, separate employers, I was urged to make a blockchain based solution to.. well actually the problem was not described in either case, only the use of blockchain.
A server whose uptime goes to X years is a server that wasn't updated in X years. I'm proud of my old XigmaNAS (FreeBSD) ~15 months uptime years ago that only a blackout did interrupt (yep, I had no UPS), and I write this as a mostly Linux user, but I'm not sure it's a good thing to aim for in production or in anything connected to the Internet.
Why do updates necessarily require reboots though?
Mine is a Debian box. Upgraded from Debian 10 to 11 to 12. Running all kinds of things. I only use Debian packages alongside one or two Docker containers.
Linux can do boring just fine.
> One of Stefano Marinelli's NetBSD boxes sat quietly serving for a decade, because everyone forgot about it. This is how Unix is meant to be. I remember someone mentioning they did this with Linux back in 1994 or 1995. Not for a decade obviously but it had been running for at least a year with no reboots or needing maintenance
There's an order of magnitude difference between running for a year and a decade... That's a lot.
It's entirely possible that BSD is more stable and lends itself better to running for really long times uninterrupted. But Linux systems running for years really aren't unheard of, and by no means is one year the top of the range.
Whether you should do that is another question. Over the course of a decade, there could well be even kernel-level vulnerabilities discovered, let alone ones in other services running on top of it. You might have a system running without a reboot for years as long as you make sure to update (and restart) user space services as needed. But leaving an entire server unattended for years doesn't sound like a good idea generally.
That may not be as much of a concern if what the box is running is a limited set of services or functionality with little exposed surface. But that then comes more down to "what you're doing with it" rather than "which OS you're using".
The generally less conservative development culture around Linux leans more towards moving fast and breaking things, although generally while trying to avoid the latter. Perhaps that makes things like low-level OS vulnerabilities or whether the system still restarts cleanly after a decade more important in the Linux land, and what counts as prudent administration in Linux might be less of a concern in BSD.
But if you can have a BSD box running for a decade, with some particular set of services, in an internal network(?), and then compare that to someone else's report of a Linux box running for (at least) a year, in 1994 or 1995, probably running an entirely different workload, in a different environment (perhaps externally exposed?), and with no indication of why it may or may not have been restarted after that time, that's not really a fair comparison either.
And hey, more power to you. But the thing is, the world has moved on. Linux is stable now. Linux containerization and VM tech is good now. The storage technologies are cross-platform now (I've had zero data loss since using zfs). The moat has shrunk.
So then, what are we left with?
* BSD is less common, so there are fewer CVEs on it. Hackers aren't stupid; they target popular platforms. You get plenty of CVEs for Apple systems (which are BSD, btw).
* Of course, since BSD is less common, it's harder to find employees who are wizards with it.
* Everyone releases for Linux first, so on some things you either have to wait, or do it yourself (or stick it in a Linux VM, at which point...)
* Licensing bites people all the time, but it's hardly a catastrophic thing; merely the normal politics that occur when things get big enough, and enough money is involved. You win by paying good money to hire the better legal team.
So what true benefit is there to using BSD over Linux? Other than satisfying your own self or going with what you're already familiar with, not much.
Earlier quoted context omitted.
I agree. I have had several services run for a decade on a company LAN using Linux. They did boring things well. They will still work in 20 years if the hardware is still there.
To be fair I've had Debian boxes with uptimes of many years and service lives of a decade. A sarge install doing nothing but DNS (on metal not VM) from 2006 lasted until 2015. So it isn't just a Linux - BSD thing it's a mindset of simplicity and focus of function. But since about 2020 my love for Debian waned, what with the systemd and ugly internal politics. Nowadays I'm inclined to look at whether a BSD or other op…
And Ubuntu really was the value-add, turnkey solution for someone who needed a desktop system as "daily driver" without endless tinkering and custom fixes.
But by 2018, Ubuntu was making enough proprietary additions that I didn't need, and I began to notice Debian's maturity and feature parity. But Debian still had a reputation for being extremely stable. So some new installs were Debian, and in 2018 when I purchased a Lenovo notebook, some were Ubuntu-certified, and I chose Red Hat certified, which ultimately ran Fedora quite smoothly.
(There were no BSD-certified notebooks.)
This veers close to treating your servers as pets instead of cattle. Which is fine if you’re small ( 99% of services are ), but not great if you have thousands of servers and scale up and down routinely.
That said, I don’t feel like that quote actually represents BSD vs Linux at all. You can have easy deploys and long term maintenance on either.
Every companies have linux servers with crazy uptimes. My server never crash, my linux desktop do. I assume they don't put BSD on their client's laptop.
If you need a stable desktop put z/OS on it, you normally don't reboot those til the end of the systems lifetime.
You mean use a mainframe remotely ?