I use Linux, FreeBSD, NetBSD, and OpenBSD all for fun, learning, and profit(the first two).
At the very least it is nice to make acquaintance with at least one BSD because it will probably expand your knowledge on Linux in ways you wont be able to anticipate.
For example, FreeBSD got me into kernel development, full system debugging, network stack development, driver development, and understanding how the whole kit fits together. Those skills transferred back and forth with reasonable fidelity to Linux, and for me jumping into Linux development cold would have been too big a leap.. especially in confidence and developing a mental model.
For my personal infrastructure, I tend to use FreeBSD because in many ways it is simpler and less surprising, especially when accounting for the passage of time. ifconfig is still ifconfig, and it works great. rc.d is all I need for my own stuff. I like the systematic effects of things like tunables and sysctl for managing hardware and kernel configuration. The man pages are forever useful to new and old users. The kernel APIs and userland APIs are extremely stable akin to commercial operating systems and unlike Linux.
There are warts. There are community frictions. The desktop story and some developer experiences will be perpetually behind Linux due to the size of the contributor base and user base. The job market for BSD is very limited compared to Linux. But I don't think it's an all or nothing affair, and ideally in a high stakes operation you would dual stack for availability and zero-day mitigation (Verisign once gave a great talk on this).
> tend to use FreeBSD because in many ways it is simpler and less surprising, especially when accounting for the passage of time. ifconfig is still ifconfig, and it works great. rc.d is all I need for my own stuff.
That sounds very appealing to me. I have to keep a small number of servers running, but its not my main focus and I would like to spend as little time on it as possible.
I have started using Alpine Linux for servers (not for my desktop, yet) because it is light and simple. Maybe BSD will be the next step.
The UI generally quite stable and well documented, which is awesome.
They also have things like `rpm` in ports that you can install. Why? Because you can enable linux binary compatibility[0] and run linux binaries on it (this implements the linux kernel interface, it's not a VM/emulator). It's also backward compatible with its own binaries back to FreeBSD 4 (circa 2000).
Though you may not need that as the ports/packages collection is pretty comprehensive.
It also comes with some nifty tools built-in for isolation (similar to but predating cgroups/containers) as "jails". It also has a hypervisor built in (bhyve) for virtualization if you do need to run any linux VMs or anything for any reason.
The way I usually sum up the difference to people is that FreeBSD is designed while Linux is grown. FreeBSD feels much more like a cohesive whole than Linux.
Really, the only reason I'm not running it everywhere is that the industry has kind of settled on linux-style containers for... absolutely everything, and the current solution for that on FreeBSD is basically "run linux in a VM".
I completely agree! If you're looking to deeply understand how Linux works under the hood, I highly recommend trying out Linux From Scratch. It gave me invaluable insight into the system, especially when I first explored it 20 years ago. Building everything from the ground up—without relying on prepackaged distros or libc—was a game changer.
I ran through that about 10 years ago. Not difficult for a reasonably seasoned Unixy person to run through but it quite frankly scared the shit out of me on the basis of how much stuff requires patching and hacking. Nothing fits together nicely. It gave me a whole new appreciation of macOS and FreeBSD.
I really really enjoyed using FreeBSD. It isn't a distro but a full OS that you can and easily fully rebuild from source. The people are great, the tech is good.
I would differ on a small point. For SOHO usage, I think that docker compose is perfectly viable and often simplifies backup, migration and moving to a new server. Just my own take on this. A lot of apps really only need one instance with a good backup strategy and not hot failover instances and can handle an hour of down time once a year or two as needed, which I rarely experience.
As mentioned in the article, it also serves as a decent set of instructions, assuming the actual dockerfile(s) for the services and dependencies are broadly available. You can swap out the compose instance of PostgreSQL for your dedicated server with a new account/db, relatively easily. Similar for other non-app centered services (redis, rabbitmq, etc). You can go all in, or partly in and in any case it does serve as self-documenting to a large degree.
> The largest failure was with btrfs — after a reboot, a 50 TB filesystem (in mirror, for backups) simply stopped working. No more mounting possible. Data was lost, but I had further backups. The client was informed and understood the situation. Within a few days, the server was rebuilt from scratch on FreeBSD with ZFS — since then, I haven’t lost a single bit.
As someone who admins a lot of btrfs, it seems very unlikely that this was unrecoverable. btrfs gets itself into scary situations, but also gets itself out again with a little effort.
In this instance “I solve problems” meant “I blow away the problem and start fresh”. Always easier! Glad the client was so understanding.
Why do people use btrfs and similar filesystems for production use? They are by no means dumpster fires. But the internet is littered with stories of "X happened, then I realized Y & that I wasn't getting my data back"
> [users ..] began explicitly requesting “jails” instead of Docker hosts. They started using BastilleBSD to clone “template” jails and deploy them.
huh, were they running persistent docker containers and modifying them in-place? If that's the case, they were missing the best part of Docker - the Dockerfile and "container is a cattle". The power of Docker is there no ad-hoc system customization possible, it's all in Dockerfile which is source-controlled and versioned, and artifacts (like built images) are read-only.
To go from this to all-manual "use bastille edit TARGET fstab to manually update the jail mounts from 13.1 to 13.2 release path." [0] seems like a real step back. I can understand why one might want to go to BSD if they prefer this kind of workflow, but for all my projects, I am now convinced that functional-like approach (or at least IaaC-like one) is much more powerful than manually editing individual files on live hosts.
I like the idea and would like to learn more, but it looks like "migrated stuff without testing ahead of time and it turned out faster for some reason". Was it the memory allocations? Was it the disk latency? The hypervisor? Could it be replicated by other means? It was a fun read, but the reasoning/understanding was missing. I hope people investigate deeper before making changes like that.
If you look for benchmarks comparing databases on Linux/BSDs you'll find lots of nuance in practice and results going both ways depending on configuration and what's being tested.
No, after 20 years of use and comparative testing of similar setups. Frankly, I have always placed little importance on benchmarks as I consider them extremely specific. I am interested in real-world use cases.
The goal of the talk and the article is not to urge people to migrate all their setups, but simply to share my experience and the results achieved. To encourage the use of BSDs for their own purposes as well. It’s not to say that they are the best solution; there is no universal solution to all problems, but having a range of choices can only be positive.
I've become a fairly loyal OpenBSD user in the last 3-4 years. The base OpenBSD load includes a substantial amount of network capabilities, and is cleanly implemented. It's almost too cleanly implemented, to the point of making me feel sort of guilty when I start to clutter up an install with a bunch of packages...
If my needs for storage were more complicated, I would probably use FreeBSD ZFS, but UFS suffices for my rather modest needs.
I use OpenBSD for desktop, web and mail services. There are some limitations, but none that are serious enough to warrant dealing with running another BSD, or Linux distribution.
I've recently discovered systemd-nspawn which is an alternative to LXC, builtin and integrated into systemd. Much lighter than full VMs and it's quite similar to Solaris Zones and FreeBSD jails. One way to use it is to extract an OCI (Docker) image to a path, that way you can reuse the container tooling provided by Docker, Podman et al.
I've barely touched the BSDs and it's been a few years since I last used Solaris so I can't make much of a comparison as a user myself.
I switched from FreeBSD to Linux, mainly because of the bad Java support and the simple fact that Linux became way more popular, which added to the difference in software availability.
FreeBSD has pretty good Java support? Sure if there is a new LTS release it takes a couple of months before it is ported to FreeBSD but that's about it?
For a long time they didn't have any. I recall from mid 90s version 4+? They were still trying to port.. or was it early 2000s lol either way it's ancient history
> I always used PostgreSQL even though most were on MySQL
That far back mysql and postgres were not even close to being a like-for-like comparison. One was a proper database, with things like referential integrity and a type system that didn't count the 31st of February to be a valid date, and the other was a glorified ndbm with some structure, a SQL interface, and was very very fast at running simple single-table SELECT statements.
Funny, my lesson was nearly the opposite: Don't try to use something that's popular but that meets my needs the best.
I used Mac for a long time but i got allergic to opinionated software. Tried Ubuntu for a while but hated the "not invented here" and commercial motives (eg even distributing small console stuff in snaps). I'm now very happy with FreeBSD on my primary desktop.
But of course my needs and things I care about are different.
You don't have to use what "most people use" if what you're using is fully compatible, or at least compatible enough that it does everything you need it to do. PostgreSQL and MySQL are both SQL databases, so depending on what exactly you're doing, there's likely not that much difference in use. (Plus, these days, PostgreSQL seems to have become more popular.)
So if you're just editing text, it doesn't matter if most people are using emacs, because you can use vim too. But I guess for running an OS, FreeBSD has had too many pain points for many people compared to just going with the Linux crowd.
You would think so, but you'd be surprised what how people (mis)use stuff and how they enjoy using vendor specific features / extensions. This is also, imo, what kind of killed the adoption of J2EE and even J2ME. I'm not talking about J2EE/JEE + Spring*, or using an embedded tomcat. I'm talking about basically what everybody's trying to create for the last 20 years in terms of deployment etc. Sun had it pretty much figured out back then
For editing text, I only use vi.. It's available everywhere and I know my way around it.
I'm glad postgres became popular :). The reasoning around using what many people use is that software is generally more available..
Maybe I'll give FreeBSD another swing in the future, but I'm not sure what the state of containers is. I used jails back in the days, but I'll have to do some research on that.
I'm interested too. I'm using XFS only, and have for many years. On my own boxes, but my company also uses XFS for all the data on customer computers. We did extensive testing many years back, and XFS was the only filesystem at the time which gave a linear, constantly very high performance when writing and reading huge amounts of data (real-time data, dips in performance is a 100% no-no), and also not degrading when having huge numbers of files. We've never had a customer lose data due to XFS problems, and at this point I can't imagine how much data that would be, except that it's astronomical.
When that's said, we had routine XFS losses on SGI boxes. That was a very well known scenario: Write constantly to a one-page text file, say, every few seconds, then power cycle the machine. The file would be empty afterwards. This doesn't happen on Linux, I vaguely recall discussing this with someone some years ago (maybe here on HN) and something was changed at some point, maybe when SGI migrated XFS to Linux, or shortly after.
It hard to know the timeline with his data loss, but I am assuming it was a long time ago.
XFS is originally from SGI Irix and was designed to run on higher end hardware. SGI donated it to Linux in 1999 and it carried a lot of its assumptions over.
For example on SGI boxes you had "hardware raid" with cache, which essentially is a sort of embedded computer with it's own memory. That cache had a battery backup so that if the machine had a crash or sudden power loss the hardware raid would live on long enough to finish its writes. SGI had tight control over the type of hardware you could use and it was usually good quality stuff.
In the land of commodity PC-based servers this isn't often how it worked. Instead you just had regular IDE or SATA hard drives. And those drives lied.
On cheap hardware the firmware would report back it had finished writes when in fact it didn't because it made it seem faster in benchmarks. And consumers/enterprise types looking to save money with Linux mostly bought whatever is the cheapest and fastest looking on benchmarks.
So that if there was a hardware failure or sudden power loss there would could be several megs of writes that were still in flight when the file system thought they were safely written to disk.
That meant there was a distinct chance of dataloss when it came to using Linux and XFS early on.
I experienced problems like that in early 2000s era Linux XFS.
This was always a big benefit to sticking with Ext4. It is kinda dumb luck that Ext4 is as fast as it is when it came to hosting databases, but the real reason to use it is because it had a lot of robust recovery tools. It was designed from the ground up with the assumption that you were using the cheapest crappiest hardware you can buy (Personal PCs).
However modern XFS is a completely different beast. It has been rewritten extensively and improved massively over what was originally ported over from SGI.
It is different enough that a guy's experience with it from 2005 or 2010 isn't really meaningful.
I have zero real technical knowledge on file systems except as a end-user, but from what I understand FreeBSD uses UFS that uses a "WAL" or "write ahead log".. where it records writes it is going to do before it does it. I think this is a simpler but more robust solution then the sort of journalling that XFS or Ext4 uses. The trade off is lower performance.
As far as ZFS vs Btrfs... I really like to avoid Btrfs as much as possible. A number of distros use it by default (OpenSUSE, Fedora, etc), but I just format everything as a single partition as Ext4 or XFS on personal stuff. I use it on my personal file server, but it really simple setup with UPS. I don't use ZFS, but I strongly suspect that btrfs simply failed to rise to its level.
One of the reasons Linux persists despite not having something up to the level of ZFS is that most of ZFS features are redundant to larger enterprise customers.
They typically use expensive SAN or more advanced NAS that has proprietary storage solutions that provide ZFS-like features long before ZFS was a thing. So throwing something as complicated as ZFS on top of that really provides no benefit.
Or they use one of Linux clustered file system solutions, of which there is a wide selection.
Facebook runs their entire stack using Btrfs [0]. I would encourage anyone who is stuck in the "oh btrfs is so buggy and loses data" mindset (not helped by articles like this [1] that play off btrfs as some half-baked contraption, when it's really btrfs raid that needs a LOT more time to bake) to look into things and realize that large companies (OpenSuse, Redhat, Faceboook) have poured a lot of time to get it to work well.
I don't know about it's multi-disk story (I do use ZFS for that personally), but for single disk options it is great. You get so many of the ZFS benefits (snapshots, rollback, easily create and delete volumes, etc) with MUCH lower memory usage (at least in my own experiments to try this out).
Facebook has stacks of thousands of spare nodes ready at any moment to replace a failed node. All essential data will be replicated across many different boxes so if a box fails you just replace it with a fresh node and replicate the data there.
This is much different to the consumer usecase where computers are pets and not cattle. A failed filesystem the night before you need to turn in your thesis may have a much larger impact on your life.
Another thing to consider is that Facebook runs btrfs on enterprise hardware (including SSDs with battery backups) which is going to be much more reliable than some chromebook which lives in the bottom of your backpack that you bring on transit every day.
Finally, I will say that the copy on write features of btrfs can result is some wildly different behaviour based upon how you use it. You can get into some very bad pathological cases with write amplification, and if you run btrfs on top of LUKS it can nearly be impossible to figure out why your disk is being pegged despite very little throughput at the VFS layer.
> I don't know about it's multi-disk story (I do use ZFS for that personally), but for single disk options it is great.
I can reliably, across vendors and drives, break RAID10 on BTRFS where MD+LVM are totally fine. Simply pull power. Discovered this when building out my latest workstation.
I haven't tried other configurations; after finding this pattern I decided to leave BTRFS for single-disk configurations where I want CoW
Meta does a lot of things that don't scale for reliable/trustworthy systems and aren't suitable for all use-cases. (I also used to work there too.)
ZFS is only reliable where it was battled-tested: on Solaris. ZoL has been infinitely tinkered with and smashed up that it's nothing like running a Thumper as a NAS.
XFS + mdadm on Linux is, without a doubt, far more reliable than ZoL. Ask me how I know. I have the scars to prove it.
Not hardly, and not in the way you think. They replaced their arguably purer ZFS port to replace it with ZoL. As such, it's nowhere near as tested and proven as existing solutions like ext4 and xfs Redhat has deployed to millions of machines for decades. ZFS has too many religious fanboys who hype it without considering that boring and reliable are less risky than betting on code that hasn't had nearly the same scale of enterprise experience.
I am stuck in the btrfs-is-buggy mindset precisely because it managed to lose my root partition on a single disk machine. It might also have raid problems, but not exclusively.
I use btrfs a lot but I'm not sure if I'd use it for production servers. The I/O bandwidth is just a lot lower and I get weird latency problems on desktop Linux when BTRFS is very busy that I don't get on other file systems. Then again, I probably wouldn't use ZFS for anything but a NAS setup either.
Yeah, and when I was there machines would run out of disk space at 50% usage and it took months to figure out why. In the mean time, they'd just reimage the machine and hope. I don't recall any issues with data loss, but it didn't have the air of reliability.
But my team was weird at FB, our uptimes of 45 days were way above the average, and we ran into all sorts of things because we operated outside the norm.
> For example on SGI boxes you had "hardware raid" with cache, which essentially is a sort of embedded computer with it's own memory. That cache had a battery backup so that if the machine had a crash or sudden power loss the hardware raid would live on long enough to finish its writes. SGI had tight control over the type of hardware you could use and it was usually good quality stuff.
Most of the SGI machine I've used of various sizes did not have hardware raid. In my experience, you were more likely to run into hardware raid on a PC than on traditional SGI or Sun servers (I don't have much experience with AIX or HP-UX), unless the unix server was in a SAN environment.
Yep. The Octanes and the Challenge servers we used at work didn't have hardware raid, and, contrary to grandparent we did have regular issues on SGI with XFS (loss of data after power cycling, always), while we never had that on Linux, which surprised me. After all, it was so easily reproduced (on SGI): Write regularly to a file, power cycle, file empty afterwards. Did that on Linux, everything fine. Every time. Never ever had losses on Linux. NB: I did not test this immediately after XFS was ported to Linux, it's very likely that things were improved on shortly after, before we started testing XFS at work.
(As for hardware raid - I didn't start to see hardware raid regularly until HP started shipping rackmount servers with Compaq raid hardware, way back. Linux boxes..)
"One of the reasons Linux persists despite not having something up to the level of ZFS is that most of ZFS features are redundant to larger enterprise customers."
ZFS is used heavily on Linux and runs well, though there are some limitations which are being addressed over time in the OpenZFS project. It is used across all areas that Linux serves, whether laptop, desktop, home server all the way to enterprise. https://openzfs.org/wiki/Main_Page
ZFS on Linux is not really usable for most users because every kernel update can break your ZFS compatibility.
Meaning unless you want to put in the time to manually test every kernel update and ensure your kernel version stays in-sync with OpenZFS you can very likely end up with an unbootable system.
> I understand FreeBSD uses UFS that uses a "WAL" or "write ahead log".. where it records writes it is going to do before it does it.
I think you're describing UFS soft-updates? I think that's more or less for meta data updates, not data data. It's been a while since I reviewed it, but it gets you nice things like snapshots and background fsck so after an unclean restart your system can get back to work immediately and clean up behind the scenes. There is some sort of journalling that's fairly new, but my experience from 10 years ago was soft-updates and background fsck just worked; and if you wanted better, ZFS was probably what you want, if you can afford copy on write.
Used it since the late 90s on IRIX, think there were a few issues early on with the endian swap, but no issues for the best part of twenty years for me!
I had one a few years back where we ran out of inodes on a Jenkins machine on CentOS 7 and it crashed and couldn’t remount the filesystem. I had to restore a backup which was time consuming on a 4TB volume with crazy amounts of files.
Indeed, most of our public-facing services are hosted on OpenBSD, and all of our routers and firewalls run it. We started managing everything with Ansible to make it easier to ignore what the host OS is for a deployment, and that has worked well both in moving things to [Open,Net]BSD for experiments, and also standing up tests on various Linux distros just to make sure we're not running into a BSD vs. GNU issue, or even a "problem only on this specific distro" issue.
#1 reason we chose Ansible over other tools was support for the BSDs.
All is fine and dandy and BSDs can solve many use cases. Unfortunately for the solution we are working on, which imply many microservices we need Kubernetes and no BSD equivalent to Kubernetes exists.
I've helped build two top 10 service provider networks (10s of Tbps). One on FreeBSD, and one on Linux with Kubernetes.
I don't really see Kubernetes as being a game changer. The biggest pro, it makes it easier to onboard both development and operations personnel having a quasi-standard for how a lot of things like scheduling and application networking work.
But it also seems to come with a magnitude of accidental and ornamental complexity. I would imply the same about microservices versus, say, figuring out your repository, language, and deployment pipelines to provide a smooth developer and operator experience. Too much of this industry is fashion and navel gazing instead of thinking about the core problems and standing behind a methodology that works for the business. Unless google moves its own infrastructure to Kubernetes, then maybe there's something to be had that couldn't reasonably be done otherwise :)
We went from a virtualized server model to managed Kubernetes and costs have escalated considerably. The additional complexity and maintenance overheads of Kubernetes are not trivial and required additional staff hires just to keep things ticking. I think the cost so far from moving from two cages in separate datacentres running blades to AWS is approximately a 6x multiplier including staff. This was all driven on the back of "we must have microservices to scale", something we have failed entirely to do. It's a complete own goal.
I think it mainly works for startups which can build it as a greenfield and are extremely sensitive to scaling (either they are exploding within 2 years or they're dead within 2 years). So they plan to scale from 0 to millions quickly. For that kinda business it makes sense because once the exploding growth comes it's super easy to scale up and even do that automatically. That's the power of Kubernetes.
For an existing business not in a hyper-growth phase with lots of restrictions and policies and processes and legacy stuff it's just extra work with no gain. It's just a round peg in a square hole. Wrong tool for the job.
Kubernetes has unquestionably been a game changer.
If you look at most enterprises today you will see it deployed everywhere.
And most of the complexity has been abstracted away by the cloud providers so all you're left with is a system that can handle all manner of different applications and deployment scenarios in a consistent way.
> Kubernetes has unquestionably been a game changer.
> If you look at most enterprises today you will see it deployed everywhere.
This doesn't mean it's a game changer. It just means it has a big cargo-cult and people keep using it regardless of whether they need it.
> most of the complexity has been abstracted away by the cloud providers
The abstractions that people add make things more complex, not less. Unless of course you don't care to understand what you're running, which is precisely the problem with the culture around most folks that use kubernetes (and more broadly the US cloud providers as an entity).
You're probably going to be downvoted to oblivion but I 100% agree with you. It seems like they tried to remove all the complexity, and just made new complexity.
FreeBSD has jails, which AFAIK are fully a match in terms of functionality, but you would need to finish implementing containerd support (maybe via runj?) and then get k8s to run on top of that. I seriously doubt there's any actual technical blockers, it's just that nobody's finished implementing all the pieces and putting them together.
"As an experiment, I decided to migrate two hosts (each with about 10 VMs) of a client — where I had full control—without telling them, over a weekend." And that's where I draw the line. Abusing the trust of your customers is an absolute no-no in my book.
I wouldn't call it abuse of trust but it's a bad idea to do a migration or any operation that can fail and cause downtime without warning the clients. Come Monday and no servers are online, what do you say, "oops, I tried to change something that didn't work"? that is fine only if they knew there was a migration over the weekend. On my end this situation would fireable offense or close to it.
Not an abuse at all. I've a contract with those clients, and I can move the VMs, change the services, etc. freely as long as it doesn't cost more than the amount we've previously set.
Otherwise, I'd never dare to do something like that.
It's still something that's weird to do without notifying the customers. What if things were slower? What if bsd introduced some slight change in behaviour that messed up their data but they didn't know when/why things changed? Full control doesn't mean unexpected YOLO changes are welcome.
As specified in several parts, the tests were conducted while maintaining and using BSD-based infrastructures for over 20 years. In some cases, Linux was used for various reasons (commercial, ideological, because they were inherited infrastructures managed by others, etc.), but the results were anticipated. I did not expect a performance degradation, and in any case, having set up the systems in a mirrored environment, there was always the possibility to revert in a few minutes.
Maybe it was all done properly. I hope so. But the post really doesn't show that which I guess is what a few people here notice. You can't say you tested something for today for over 20 years. Things change - if it wasn't tested for that specific migration for that specific customer, then it wasn't really tested. I see people doing yolo changes that way and thought it's worth mentioning explicitly.
What!? Changing implementation details is not “abusing trust”. Where would you even draw the line with this attitude!? Should I be informing my customers whenever I update the version of left-pad I have installed!?
They're paying for a service. For example, if their Nextcloud is working and is stable, they don't care if it's running on Linux, FreeBSD, OmniOS, etc.
It's in the contract we have - and they're fine with it.
> Abusing the trust of your customers is an absolute no-no in my book.
How do people on the Internet come to such random conclusions when there is no way you could have known the full terms of the contract between the author and their client?
Abusing trust is a bit strong, customers pay for a service and beyond a certain level of abstraction these obscure technical details (from their perspective) are not their concern. They're paying to have that abstracted.
How is it an abuse? As long as the customer continues to receive the service they paid for, who cares?
The major providers such as GCP, AWS etc share very few details about their underlying infrastructure with their customers. They change all sorts of things all the time.
If I want to run a BSD as just a file server - so I guess zfs + samba + bonjour or whatever the discovery protocol is these days - which BSD should I try?
I'd recommend using FreeBSD as your first BSD. It has a more recent version of ZFS integrated in the kernel than NetBSD. OpenBSD does not have ZFS support; it's a direction they chose not to take for security and simplicity reasons.
All I can say is experiences differ. I'm a long time Debian user, and now use FreeBSD for work. Both are far better than the proprietary competition, but I'd take Debian/Linux over FreeBSD when building a random server.
To give but one example, I recently reported a bug when FreeBSD didn't boot after upgrade from 13 to 14. Worse the disk format was somehow altered so when the reboot tried to boot off 13 due to zfs bootonce flag (supposedly a failsafe), it refused to boot for the same reason. I believe it's due to a race condition in geom/cam. The same symptoms were reported 6 years ago, but the bug report has seen no activity. Making your system irrecoverable without a rescue image and console access strikes me as pretty serious. He waxes lyrical about zfs, but it's slower and more resource hungry than it's simpler competition and it's not difficult to find numerous serious zfs bug reports over the years. (But not slower than FreeBSD's UFS, oddly. It's impressively slow.) Another thing that sticks in my mind is a core zfs contributor saying it's encryption support should never have been merged.
This sounds too disparaging because the simplicity and size of FreeBSD has its own charms, but the "it's all sunshine and roses" picture he paints doesn't ring true to me. While it's probably fair to say stable versions of FreeBSD are better than the Linux kernels from kernel.org, and possibly Fedora and Ubuntu, they definitely trail behind the standard Debian stable releases.
Comparing FreeBSD to Debian makes throws up some interesting tradeoffs. On the one hand, FreeBSD's init system is a delight compared to systemd. Sure, systemd can do far, far more. But that added complexity makes for a steep learning curve and a lot of weird and difficult to debug problems, and as FreeBSD's drop dead simple /etc/rc.conf system proves most of the time the complexity isn't needed to get the job done. FreeBSD's jails just make more intuitive sense to me than Linux's equivalent which is built on control groups. FreeBSD's source is a joy to read compared to most I've seen elsewhere. I don't know who's responsible for that - but they deserve a medal.
On the downside - what were they thinking when they came up with the naming scheme under /dev for block devices? (Who thought withering device names was a good idea, so that /dev no longer reflects the state of attached hardware?) And a piece of free advice - just copy how Linux has does it's booting. Loading a kernel + initramfs is both simpler and far more flexible then the FreeBSD loader scheme. Hell, it's so flexible you can replace a BIOS with it.
The combination of the best parts Linux and the BSD's would make for an wonderful world. But having a healthy selection of choices is probably more important, and yes - I agree with him that if you are building an appliance that has an OS embedded in it, the simplicity of FreeBSD does give it an edge.
I still remember discovering FidoNet sometime in the 90s.
It was a time where sending regular mail to different countries could take weeks, and cross country phonecalls would cost between $2 and $20 per minute, and here was FidoNet that promised to allow communications across the globe with only 1-4 days delay and basically for free.
My 15-18 year old self was instantly sold. I spent countless hours reading through the "forums" on there. So much knowledge just at the tip of my fingers.
Of course some time later it was more or less replaced (for me) by email, usenet and IRC, but the memory still remains.
... This site and this directory were
driving up server load averages, which causes instability for all users on
this server. Our block on this site appears to have blocked over 150,000
connections in about half an hour. This is simply more than is appropriate
for a shared server.
You may want to review your domain's traffic logs to see what kind of
traffic this site is getting. The logs for today will be distributed at
midnight EST tonight. If the traffic to this site is legitimate, you
should look into a dedicated or a virtual private server. If the traffic
is not legitimate, you should block it. I can provide assistance with that
if you need it.
This domain and this directory will need to be disabled until traffic dies
down, traffic is blocked, or your account is upgraded from shared hosting.
If you need access to the directory so that you can make any necessary
changes, please let me know.
Sure, it's reasonable, if your desired outcome is to lose a customer by
a) demonstrating a lack of burst or headroom capacity and/or graceful degradation/priority scheduling capability, and
b) being a dick about it.
Alternatively one may intervene with a positive approach and framing viz. "we have temporarily [provisioned additional capacity / relocated your storage to another spindle / mirrored your content to a less congested host / etc] in order to handle your traffic, can we call you tomorrow to move you to a service plan that can handle your growth". Even just throttling the account in some appropriate fashion would've been better than the blunt "we dropped you like a hot potato" that was actually communicated. This was in 2010, by which time any ISP that wasn't operating as amateur hour had the tools for all these things.
> This was in 2010, by which time any ISP that wasn't operating as amateur hour had the tools for all these things.
Agreed. Also consider that shared hosting packages are very much providers scraping the bottom of the barrel for some marginal extra revenue. Shared hosting customers are not the highest priority of customers, unless your entire business is based around shared hosting (in which case, you are likely operating as amateur hour anyway)
depends on what product/service plan they had. a 5$ per month is not worth jumping through the hoops, because usually an individual solution is needed, but this doesn't scale. if the site is important to the owner, they need a more fitting service plan with higher headroom.
that being said, a good provider at least tries to filter out what could be a (d)dos
In 2010, those tools were not easily available outside massive Enterprise solutions like Akamai. Cloudflare just launched in 2009 and on-premise hardware was generally cost prohibitive.
In many industries it's standard practice to cause pain to one customer to avoid causing pain to all of them. If you've ever had work done on your house, a small guy who overruns will often bail out on your job ( and reschedule) rather than slip all the other customers he had lined up. so I can see why this guy decided to disable a directory rather than all the other customers in the server have slow service
The issue is that the economics are different: a roofer doesn't earn more by doing one big job than 4 small ones in the same time. But a big user is a better economic prospect to a service provider than the small ones. That's why this was clearly a dumb move
They should have been monitoring load and notified the customer before it got to the point that they were worried about stability.
Shutting down a customer's site without warning when you could have notified them of a problem in the near future... well, that's a great way to lose a customer. I really hope YC promptly moved to another provider after that incident.
If this was their first and only mail on that yeah it's not the best way to handle. But then my sites don't generate anywhere near the traffic as YC even YC of 2010
All a man needs is food in his stomach and a place to rest at the end of the day. Everything else is vanity
What proportion of global GDP is dedicated to fulfilling our basic material needs?
It is mostly unnecessary. Inspite of the huge productivity gains made since the seventies, the current generation of young Americans are poorer than their parents and grandparents were at their age.
So what does all the IT optimization bring? Just more wealth for the owners and redundancies for their employees, including Joe Bloggs here.
It is time people in IT got to understand this. In the long term their activities are not going to improve their wealth. They are one of the few professions whose job is to optimize themselves out of a living, unless they own the means of the production their are optimizing, which they don't.
The current generation of Americans are absolutely rich enough to just get food and have shelter. The ones that struggle are the ones that want to live in popular cities precisely because of the available improvements beyond shelter and food.
The houses of the 50s were shit tier and spread around the entire US. You can go buy them today for cheap in the 98% of locations people don’t want to live in.
He only works for himself only if he gets a steady income stream so long as the systems he manages stay up, which I'm sure he doesn't.
There is a reason for the "you will own nothing and you will be happy" ideology being promoted by "PTB", ie subscriptions for ink cartridges, heated seats and advanced suspensions in newer cars.
Corporations now want a continuous income stream from the services provide by physical products they have "sold", but they don't want their employees and subcontractors from earning some of that income stream.
Some IT administrators have been know to schedule regular "downtimes" on perfectly performing systems just to ensure their users and bosses don't take their service for granted.
I have kind of scratched my head a bit as to why our local VM provider has all these scheduled downtimes.. while the servers we run in-house have those maybe once or twice a year, at most. Hm..
> why our local VM provider has all these scheduled downtimes
Generally (but not exclusively) this is because they have more complex infrastructure, a larger number of customers, and a contractual commitment to give advance notice of outages when possible.
It therefore makes sense to schedule downtime even if you don't need it (or even use it), so that customers can arrange their own expectations.
> There is a reason for the "you will own nothing and you will be happy" ideology being promoted by "PTB",
No one, and certainly not the "PTB", is promoting this idea.
It was a thought experiment in a book/TED talk that deeply offends a certain percentage of the populace, and is ignored by everyone else. It is nothing more than that.
It does not exist in any discussions outside the circles of those who are offended for profit, and those that listen to them.
Spare yourself (and us), and stop listening to them.
I really wanted to give a try to FreeBSD... because I thought it was linux from "better" times... and then I saw they are tied to the tons of similar gcc extensions... and then I say to myself "why bother since it has the same major compiler dependency issue", better try to fix linux code base or start from there.
Please don't do this here. Not that we don't appreciate a good old-fashioned flame, but that the long-term costs outweigh the benefits, and we want this site to survive long-term.
I've a contract with those clients, and I can move the VMs, change the services, etc. freely as long as it doesn't cost more than the amount we've previously set.
Otherwise, I'd never dare to do something like that.
And I'm not so crazy as to do such an operation without the appropriate tests and foundations. Of course, when I started, I had all the conditions to be able to do it, and I had already conducted all possible tests. :-)
The client is paying for the VM. The underlying system is an abstraction. As long as service agreements weren’t interrupted I don’t see the problem. It sounds shady to say “without telling them,” because saying so implies they should have. I do a lot of optimizations for my customers without telling them, it’s not usually worth mentioning. I assume what they intended to convey was that this change caused no interruption of service so there was no need to contact or warn the customer.
This is similar to AWS S3 object storage -- AWS has over the years changed how they store their S3 data -- however as long as the API responds the same way every time it's all good. Personally I would probably do some A/B testing -- migrate half the workload and compare A to B to see if the new system is performing better before migrating the other half.
No, it's not. S3 has a very well defined API with easily measureable performance parameters. So AWS updates can make sure they don't make things worse.
This is not possible with a client's workload unless you can actually test it. That's why AWS will warn you multiple times if they need to migrate your EC2 instance onto a different hardware node. Even if it is technically "better".
Of course, the fact that clients trust their workloads to this guy probably means that there was nothing important there.
The author also converted some of these VMs to jails, so I assume they have root on the VMs (and the customers want them to admin the host). That means they should be able to see the application level performance metrics.
The workload got several times faster. The customer’s only concern was that they might be accidentally running on a more expensive instance.
In every system I’ve worked on, the agreement is in terms of an SLO. We never gave our customers any sort of expectation (or guarantee) that we wouldn’t suddenly wildly beat our SLO targets (and, in fact, we often did, due to routine upgrades).
Having said that, certain customers dictate production freezes during launches, or only want to run stuff that’s been baked in production elsewhere for 3-6 months. Upgrading those customers behind their backs would be unacceptable, especially because they pay extra for a crappier but more stable setup.
At the very least it is nice to make acquaintance with at least one BSD because it will probably expand your knowledge on Linux in ways you wont be able to anticipate.
For example, FreeBSD got me into kernel development, full system debugging, network stack development, driver development, and understanding how the whole kit fits together. Those skills transferred back and forth with reasonable fidelity to Linux, and for me jumping into Linux development cold would have been too big a leap.. especially in confidence and developing a mental model.
For my personal infrastructure, I tend to use FreeBSD because in many ways it is simpler and less surprising, especially when accounting for the passage of time. ifconfig is still ifconfig, and it works great. rc.d is all I need for my own stuff. I like the systematic effects of things like tunables and sysctl for managing hardware and kernel configuration. The man pages are forever useful to new and old users. The kernel APIs and userland APIs are extremely stable akin to commercial operating systems and unlike Linux.
There are warts. There are community frictions. The desktop story and some developer experiences will be perpetually behind Linux due to the size of the contributor base and user base. The job market for BSD is very limited compared to Linux. But I don't think it's an all or nothing affair, and ideally in a high stakes operation you would dual stack for availability and zero-day mitigation (Verisign once gave a great talk on this).
That sounds very appealing to me. I have to keep a small number of servers running, but its not my main focus and I would like to spend as little time on it as possible.
I have started using Alpine Linux for servers (not for my desktop, yet) because it is light and simple. Maybe BSD will be the next step.
They also have things like `rpm` in ports that you can install. Why? Because you can enable linux binary compatibility[0] and run linux binaries on it (this implements the linux kernel interface, it's not a VM/emulator). It's also backward compatible with its own binaries back to FreeBSD 4 (circa 2000).
Though you may not need that as the ports/packages collection is pretty comprehensive.
It also comes with some nifty tools built-in for isolation (similar to but predating cgroups/containers) as "jails". It also has a hypervisor built in (bhyve) for virtualization if you do need to run any linux VMs or anything for any reason.
The way I usually sum up the difference to people is that FreeBSD is designed while Linux is grown. FreeBSD feels much more like a cohesive whole than Linux.
Really, the only reason I'm not running it everywhere is that the industry has kind of settled on linux-style containers for... absolutely everything, and the current solution for that on FreeBSD is basically "run linux in a VM".
[0] https://docs.freebsd.org/en/books/handbook/linuxemu/
Check it out: https://www.linuxfromscratch.org/
The CON is coming up, https://freebsdfoundation.org/news-and-events/event-calendar...
Fall 2024 FreeBSD Summit November 7-8, 2024 San Jose, CA
As mentioned in the article, it also serves as a decent set of instructions, assuming the actual dockerfile(s) for the services and dependencies are broadly available. You can swap out the compose instance of PostgreSQL for your dedicated server with a new account/db, relatively easily. Similar for other non-app centered services (redis, rabbitmq, etc). You can go all in, or partly in and in any case it does serve as self-documenting to a large degree.
As someone who admins a lot of btrfs, it seems very unlikely that this was unrecoverable. btrfs gets itself into scary situations, but also gets itself out again with a little effort.
In this instance “I solve problems” meant “I blow away the problem and start fresh”. Always easier! Glad the client was so understanding.
As someone who used it all day every day in my day job for 4 years, I find it 100% believable.
I am not saying you're wrong: I'm saying, experiences differ widely, and your patterns of use are not be universal.
It's the single most unreliable untrustworthy filesystem I've used in the 21st century.
huh, were they running persistent docker containers and modifying them in-place? If that's the case, they were missing the best part of Docker - the Dockerfile and "container is a cattle". The power of Docker is there no ad-hoc system customization possible, it's all in Dockerfile which is source-controlled and versioned, and artifacts (like built images) are read-only.
To go from this to all-manual "use bastille edit TARGET fstab to manually update the jail mounts from 13.1 to 13.2 release path." [0] seems like a real step back. I can understand why one might want to go to BSD if they prefer this kind of workflow, but for all my projects, I am now convinced that functional-like approach (or at least IaaC-like one) is much more powerful than manually editing individual files on live hosts.
[0] https://bastille.readthedocs.io/en/latest/chapters/upgrading...
If you look for benchmarks comparing databases on Linux/BSDs you'll find lots of nuance in practice and results going both ways depending on configuration and what's being tested.
The goal of the talk and the article is not to urge people to migrate all their setups, but simply to share my experience and the results achieved. To encourage the use of BSDs for their own purposes as well. It’s not to say that they are the best solution; there is no universal solution to all problems, but having a range of choices can only be positive.
If my needs for storage were more complicated, I would probably use FreeBSD ZFS, but UFS suffices for my rather modest needs.
I use OpenBSD for desktop, web and mail services. There are some limitations, but none that are serious enough to warrant dealing with running another BSD, or Linux distribution.
I've barely touched the BSDs and it's been a few years since I last used Solaris so I can't make much of a comparison as a user myself.
My lesson was to just use what moet people use. But then again, I always used PostgreSQL even though most were on MySQL
That far back mysql and postgres were not even close to being a like-for-like comparison. One was a proper database, with things like referential integrity and a type system that didn't count the 31st of February to be a valid date, and the other was a glorified ndbm with some structure, a SQL interface, and was very very fast at running simple single-table SELECT statements.
Because they would just use a SQL library or ORM which hid all of these details.
Funny, my lesson was nearly the opposite: Don't try to use something that's popular but that meets my needs the best.
I used Mac for a long time but i got allergic to opinionated software. Tried Ubuntu for a while but hated the "not invented here" and commercial motives (eg even distributing small console stuff in snaps). I'm now very happy with FreeBSD on my primary desktop.
But of course my needs and things I care about are different.
So if you're just editing text, it doesn't matter if most people are using emacs, because you can use vim too. But I guess for running an OS, FreeBSD has had too many pain points for many people compared to just going with the Linux crowd.
For editing text, I only use vi.. It's available everywhere and I know my way around it.
I'm glad postgres became popular :). The reasoning around using what many people use is that software is generally more available..
Maybe I'll give FreeBSD another swing in the future, but I'm not sure what the state of containers is. I used jails back in the days, but I'll have to do some research on that.
I also recall a problem with mmap() but I’m not sure if it was related to Java or something else
When that's said, we had routine XFS losses on SGI boxes. That was a very well known scenario: Write constantly to a one-page text file, say, every few seconds, then power cycle the machine. The file would be empty afterwards. This doesn't happen on Linux, I vaguely recall discussing this with someone some years ago (maybe here on HN) and something was changed at some point, maybe when SGI migrated XFS to Linux, or shortly after.
XFS is originally from SGI Irix and was designed to run on higher end hardware. SGI donated it to Linux in 1999 and it carried a lot of its assumptions over.
For example on SGI boxes you had "hardware raid" with cache, which essentially is a sort of embedded computer with it's own memory. That cache had a battery backup so that if the machine had a crash or sudden power loss the hardware raid would live on long enough to finish its writes. SGI had tight control over the type of hardware you could use and it was usually good quality stuff.
In the land of commodity PC-based servers this isn't often how it worked. Instead you just had regular IDE or SATA hard drives. And those drives lied.
On cheap hardware the firmware would report back it had finished writes when in fact it didn't because it made it seem faster in benchmarks. And consumers/enterprise types looking to save money with Linux mostly bought whatever is the cheapest and fastest looking on benchmarks.
So that if there was a hardware failure or sudden power loss there would could be several megs of writes that were still in flight when the file system thought they were safely written to disk.
That meant there was a distinct chance of dataloss when it came to using Linux and XFS early on.
I experienced problems like that in early 2000s era Linux XFS.
This was always a big benefit to sticking with Ext4. It is kinda dumb luck that Ext4 is as fast as it is when it came to hosting databases, but the real reason to use it is because it had a lot of robust recovery tools. It was designed from the ground up with the assumption that you were using the cheapest crappiest hardware you can buy (Personal PCs).
However modern XFS is a completely different beast. It has been rewritten extensively and improved massively over what was originally ported over from SGI.
It is different enough that a guy's experience with it from 2005 or 2010 isn't really meaningful.
I have zero real technical knowledge on file systems except as a end-user, but from what I understand FreeBSD uses UFS that uses a "WAL" or "write ahead log".. where it records writes it is going to do before it does it. I think this is a simpler but more robust solution then the sort of journalling that XFS or Ext4 uses. The trade off is lower performance.
As far as ZFS vs Btrfs... I really like to avoid Btrfs as much as possible. A number of distros use it by default (OpenSUSE, Fedora, etc), but I just format everything as a single partition as Ext4 or XFS on personal stuff. I use it on my personal file server, but it really simple setup with UPS. I don't use ZFS, but I strongly suspect that btrfs simply failed to rise to its level.
One of the reasons Linux persists despite not having something up to the level of ZFS is that most of ZFS features are redundant to larger enterprise customers.
They typically use expensive SAN or more advanced NAS that has proprietary storage solutions that provide ZFS-like features long before ZFS was a thing. So throwing something as complicated as ZFS on top of that really provides no benefit.
Or they use one of Linux clustered file system solutions, of which there is a wide selection.
I don't know about it's multi-disk story (I do use ZFS for that personally), but for single disk options it is great. You get so many of the ZFS benefits (snapshots, rollback, easily create and delete volumes, etc) with MUCH lower memory usage (at least in my own experiments to try this out).
[0] https://lwn.net/Articles/824855/ [1] https://arstechnica.com/gadgets/2021/09/examining-btrfs-linu...
This is much different to the consumer usecase where computers are pets and not cattle. A failed filesystem the night before you need to turn in your thesis may have a much larger impact on your life.
Another thing to consider is that Facebook runs btrfs on enterprise hardware (including SSDs with battery backups) which is going to be much more reliable than some chromebook which lives in the bottom of your backpack that you bring on transit every day.
Finally, I will say that the copy on write features of btrfs can result is some wildly different behaviour based upon how you use it. You can get into some very bad pathological cases with write amplification, and if you run btrfs on top of LUKS it can nearly be impossible to figure out why your disk is being pegged despite very little throughput at the VFS layer.
I can reliably, across vendors and drives, break RAID10 on BTRFS where MD+LVM are totally fine. Simply pull power. Discovered this when building out my latest workstation.
I haven't tried other configurations; after finding this pattern I decided to leave BTRFS for single-disk configurations where I want CoW
ZFS is only reliable where it was battled-tested: on Solaris. ZoL has been infinitely tinkered with and smashed up that it's nothing like running a Thumper as a NAS.
XFS + mdadm on Linux is, without a doubt, far more reliable than ZoL. Ask me how I know. I have the scars to prove it.
I also owned a Thumper and Thor running Solaris in 2009. Much prefer Linux and the hardware solutions today.
What specific problems are there with the ZFS implementation on FreeBSD? You claim it is not battle tested, I find it to be rock solid..
Those two things are not mutually exclusive.
The cause? Filling up the filesystem. Why? Because of OS snapshots.
(Aside: why can they fill it? Because it doesn't give a straight answer to `df -h`. Why not? Because of snapshots.)
Yeah, and when I was there machines would run out of disk space at 50% usage and it took months to figure out why. In the mean time, they'd just reimage the machine and hope. I don't recall any issues with data loss, but it didn't have the air of reliability.
But my team was weird at FB, our uptimes of 45 days were way above the average, and we ran into all sorts of things because we operated outside the norm.
Edit: it's less grey now.
Most of the SGI machine I've used of various sizes did not have hardware raid. In my experience, you were more likely to run into hardware raid on a PC than on traditional SGI or Sun servers (I don't have much experience with AIX or HP-UX), unless the unix server was in a SAN environment.
(As for hardware raid - I didn't start to see hardware raid regularly until HP started shipping rackmount servers with Compaq raid hardware, way back. Linux boxes..)
ZFS is used heavily on Linux and runs well, though there are some limitations which are being addressed over time in the OpenZFS project. It is used across all areas that Linux serves, whether laptop, desktop, home server all the way to enterprise. https://openzfs.org/wiki/Main_Page
Meaning unless you want to put in the time to manually test every kernel update and ensure your kernel version stays in-sync with OpenZFS you can very likely end up with an unbootable system.
https://kernel.org/
I think you're describing UFS soft-updates? I think that's more or less for meta data updates, not data data. It's been a while since I reviewed it, but it gets you nice things like snapshots and background fsck so after an unclean restart your system can get back to work immediately and clean up behind the scenes. There is some sort of journalling that's fairly new, but my experience from 10 years ago was soft-updates and background fsck just worked; and if you wanted better, ZFS was probably what you want, if you can afford copy on write.
#1 reason we chose Ansible over other tools was support for the BSDs.
I don't really see Kubernetes as being a game changer. The biggest pro, it makes it easier to onboard both development and operations personnel having a quasi-standard for how a lot of things like scheduling and application networking work.
But it also seems to come with a magnitude of accidental and ornamental complexity. I would imply the same about microservices versus, say, figuring out your repository, language, and deployment pipelines to provide a smooth developer and operator experience. Too much of this industry is fashion and navel gazing instead of thinking about the core problems and standing behind a methodology that works for the business. Unless google moves its own infrastructure to Kubernetes, then maybe there's something to be had that couldn't reasonably be done otherwise :)
We went from a virtualized server model to managed Kubernetes and costs have escalated considerably. The additional complexity and maintenance overheads of Kubernetes are not trivial and required additional staff hires just to keep things ticking. I think the cost so far from moving from two cages in separate datacentres running blades to AWS is approximately a 6x multiplier including staff. This was all driven on the back of "we must have microservices to scale", something we have failed entirely to do. It's a complete own goal.
For an existing business not in a hyper-growth phase with lots of restrictions and policies and processes and legacy stuff it's just extra work with no gain. It's just a round peg in a square hole. Wrong tool for the job.
You could have easily run Kubernetes on your virtualised servers.
If you look at most enterprises today you will see it deployed everywhere.
And most of the complexity has been abstracted away by the cloud providers so all you're left with is a system that can handle all manner of different applications and deployment scenarios in a consistent way.
> If you look at most enterprises today you will see it deployed everywhere.
This doesn't mean it's a game changer. It just means it has a big cargo-cult and people keep using it regardless of whether they need it.
> most of the complexity has been abstracted away by the cloud providers
The abstractions that people add make things more complex, not less. Unless of course you don't care to understand what you're running, which is precisely the problem with the culture around most folks that use kubernetes (and more broadly the US cloud providers as an entity).
Otherwise, I'd never dare to do something like that.
Its also a legal nightmare for the hoster if something goes wrong.
How do people on the Internet come to such random conclusions when there is no way you could have known the full terms of the contract between the author and their client?
Yes. I also always let my customers sign off when I change the libraries I use. Completely sane approach.
The major providers such as GCP, AWS etc share very few details about their underlying infrastructure with their customers. They change all sorts of things all the time.
To give but one example, I recently reported a bug when FreeBSD didn't boot after upgrade from 13 to 14. Worse the disk format was somehow altered so when the reboot tried to boot off 13 due to zfs bootonce flag (supposedly a failsafe), it refused to boot for the same reason. I believe it's due to a race condition in geom/cam. The same symptoms were reported 6 years ago, but the bug report has seen no activity. Making your system irrecoverable without a rescue image and console access strikes me as pretty serious. He waxes lyrical about zfs, but it's slower and more resource hungry than it's simpler competition and it's not difficult to find numerous serious zfs bug reports over the years. (But not slower than FreeBSD's UFS, oddly. It's impressively slow.) Another thing that sticks in my mind is a core zfs contributor saying it's encryption support should never have been merged.
This sounds too disparaging because the simplicity and size of FreeBSD has its own charms, but the "it's all sunshine and roses" picture he paints doesn't ring true to me. While it's probably fair to say stable versions of FreeBSD are better than the Linux kernels from kernel.org, and possibly Fedora and Ubuntu, they definitely trail behind the standard Debian stable releases.
Comparing FreeBSD to Debian makes throws up some interesting tradeoffs. On the one hand, FreeBSD's init system is a delight compared to systemd. Sure, systemd can do far, far more. But that added complexity makes for a steep learning curve and a lot of weird and difficult to debug problems, and as FreeBSD's drop dead simple /etc/rc.conf system proves most of the time the complexity isn't needed to get the job done. FreeBSD's jails just make more intuitive sense to me than Linux's equivalent which is built on control groups. FreeBSD's source is a joy to read compared to most I've seen elsewhere. I don't know who's responsible for that - but they deserve a medal.
On the downside - what were they thinking when they came up with the naming scheme under /dev for block devices? (Who thought withering device names was a good idea, so that /dev no longer reflects the state of attached hardware?) And a piece of free advice - just copy how Linux has does it's booting. Loading a kernel + initramfs is both simpler and far more flexible then the FreeBSD loader scheme. Hell, it's so flexible you can replace a BIOS with it.
The combination of the best parts Linux and the BSD's would make for an wonderful world. But having a healthy selection of choices is probably more important, and yes - I agree with him that if you are building an appliance that has an OS embedded in it, the simplicity of FreeBSD does give it an edge.
Sorry, could you clarify what this means? I'm not super familiar with freebsd and don't understand what withering means here.
It was a time where sending regular mail to different countries could take weeks, and cross country phonecalls would cost between $2 and $20 per minute, and here was FidoNet that promised to allow communications across the globe with only 1-4 days delay and basically for free.
My 15-18 year old self was instantly sold. I spent countless hours reading through the "forums" on there. So much knowledge just at the tip of my fingers.
Of course some time later it was more or less replaced (for me) by email, usenet and IRC, but the memory still remains.
http://www.archub.org/ycoutage.txt
... This site and this directory were driving up server load averages, which causes instability for all users on this server. Our block on this site appears to have blocked over 150,000 connections in about half an hour. This is simply more than is appropriate for a shared server.
You may want to review your domain's traffic logs to see what kind of traffic this site is getting. The logs for today will be distributed at midnight EST tonight. If the traffic to this site is legitimate, you should look into a dedicated or a virtual private server. If the traffic is not legitimate, you should block it. I can provide assistance with that if you need it.
This domain and this directory will need to be disabled until traffic dies down, traffic is blocked, or your account is upgraded from shared hosting. If you need access to the directory so that you can make any necessary changes, please let me know.
a) demonstrating a lack of burst or headroom capacity and/or graceful degradation/priority scheduling capability, and
b) being a dick about it.
Alternatively one may intervene with a positive approach and framing viz. "we have temporarily [provisioned additional capacity / relocated your storage to another spindle / mirrored your content to a less congested host / etc] in order to handle your traffic, can we call you tomorrow to move you to a service plan that can handle your growth". Even just throttling the account in some appropriate fashion would've been better than the blunt "we dropped you like a hot potato" that was actually communicated. This was in 2010, by which time any ISP that wasn't operating as amateur hour had the tools for all these things.
Agreed. Also consider that shared hosting packages are very much providers scraping the bottom of the barrel for some marginal extra revenue. Shared hosting customers are not the highest priority of customers, unless your entire business is based around shared hosting (in which case, you are likely operating as amateur hour anyway)
The issue is that the economics are different: a roofer doesn't earn more by doing one big job than 4 small ones in the same time. But a big user is a better economic prospect to a service provider than the small ones. That's why this was clearly a dumb move
They should have been monitoring load and notified the customer before it got to the point that they were worried about stability.
Shutting down a customer's site without warning when you could have notified them of a problem in the near future... well, that's a great way to lose a customer. I really hope YC promptly moved to another provider after that incident.
Bloke is not acquainted with Keynesian economics.
https://www.youtube.com/watch?v=9OhIdDNtSv0
https://www.youtube.com/watch?v=NO_tTnpof_o
All a man needs is food in his stomach and a place to rest at the end of the day. Everything else is vanity
What proportion of global GDP is dedicated to fulfilling our basic material needs?
It is mostly unnecessary. Inspite of the huge productivity gains made since the seventies, the current generation of young Americans are poorer than their parents and grandparents were at their age.
So what does all the IT optimization bring? Just more wealth for the owners and redundancies for their employees, including Joe Bloggs here.
It is time people in IT got to understand this. In the long term their activities are not going to improve their wealth. They are one of the few professions whose job is to optimize themselves out of a living, unless they own the means of the production their are optimizing, which they don't.
It is their employers that do.
The houses of the 50s were shit tier and spread around the entire US. You can go buy them today for cheap in the 98% of locations people don’t want to live in.
Sounds like he understood it just fine. He owns the means of production.
There is a reason for the "you will own nothing and you will be happy" ideology being promoted by "PTB", ie subscriptions for ink cartridges, heated seats and advanced suspensions in newer cars.
Corporations now want a continuous income stream from the services provide by physical products they have "sold", but they don't want their employees and subcontractors from earning some of that income stream.
Some IT administrators have been know to schedule regular "downtimes" on perfectly performing systems just to ensure their users and bosses don't take their service for granted.
Generally (but not exclusively) this is because they have more complex infrastructure, a larger number of customers, and a contractual commitment to give advance notice of outages when possible.
It therefore makes sense to schedule downtime even if you don't need it (or even use it), so that customers can arrange their own expectations.
No one, and certainly not the "PTB", is promoting this idea.
It was a thought experiment in a book/TED talk that deeply offends a certain percentage of the populace, and is ignored by everyone else. It is nothing more than that.
It does not exist in any discussions outside the circles of those who are offended for profit, and those that listen to them.
Spare yourself (and us), and stop listening to them.
and that should be the title of this post too.
I like that the blog post shares the slides, not just the video.
Otherwise, I'd never dare to do something like that.
And I'm not so crazy as to do such an operation without the appropriate tests and foundations. Of course, when I started, I had all the conditions to be able to do it, and I had already conducted all possible tests. :-)
This is not possible with a client's workload unless you can actually test it. That's why AWS will warn you multiple times if they need to migrate your EC2 instance onto a different hardware node. Even if it is technically "better".
Of course, the fact that clients trust their workloads to this guy probably means that there was nothing important there.
The VM change was sufficient enough to alter the runtime of a task by several times. This is NOT a small inconsequential change.
You _have_ to warn your clients when you do stuff like this.
In every system I’ve worked on, the agreement is in terms of an SLO. We never gave our customers any sort of expectation (or guarantee) that we wouldn’t suddenly wildly beat our SLO targets (and, in fact, we often did, due to routine upgrades).
Having said that, certain customers dictate production freezes during launches, or only want to run stuff that’s been baked in production elsewhere for 3-6 months. Upgrading those customers behind their backs would be unacceptable, especially because they pay extra for a crappier but more stable setup.
And he found it out only by talking to the customer, indicating approximately zero testing from his side.
It might have gone the other way easily.