btrfs: The Best Filesystem You've Never Heard Of

What? You've never heard about btrfs? Well, today's your lucky day! Come and hear about this relatively new filesystem, why it's awesome, and learn why it's been called out in poiupoiu's bio for two years now! t? You've never heard about btrfs? Well, today's your lucky day! Come and hear about this relatively new filesystem, why it's awesome, and learn why it's been called out in poiupoiu's bio for two years now!
About Poiupoiu:
Poiupoiu is a daddy, a husband, an engineer, and a geek from Huntsville, AL. He enjoys encoding videos, photography, playing video games, and puns. He has spent countless hours learning about video encoding, and loves him some SSH, rsync, and btrfs. He has been a great friend to PhreakNIC and serves as our Audio/Visual Director this year.

[@pezilord]

I ended up here trying to find an olive bread recipe.

[@nextlifeonearth]

I've got a sourdough recipe with olives and rosemary. 500g flour (I use 300g white, 50g rye, 150g whole wheat, but you can mix whatever) 200g sourdough starter 280g water (as you noticed, I weigh everything including the liquids, in the bowl) 8g salt 10g extra virgin olive oil 10g olives (no pits, cut into 2-4 mm slices) "about yay much" rosemary. (to taste) Mix flour, water, sourdough, salt and oil. Rise for at least 6 hours in fridge or 4 outside (sourdough takes a long time. Best is to do this in the morning for the next day, put it in the fridge in the morning for the next step that evening) Add olives and thyme, stretch and fold it into the dough. Rise for at least another 6 hours in fridge again (overnight is easiest. Longer up to a week is better flavour.) Preheat oven at 250 C (depends on your oven) Leave a heat resistant bowl in the bottom. Boil water on the side. Shape the bread however, but do it whilst developing the gluten for a firmer bread. Bake for 35 minutes (depends on oven again. Test after 30 minutes with a wooden skewer. If it comes out clean-ish, you can take it out.) Let the bread cool and do its baking in residual heat under a towel for half an hour or so. Should have a pretty good bread now.

[@robinmattheussen2395]

Stratis is partially developed in Python because its CLI is written in it. I actually guessed that before looking it up. Maybe check that out before making those comments about it? Also, Stratis is not a filesystem, it's just a daemon that can be used to configure the underlying subsystems in a way that could potentially give us similar features to btrfs or ZFS. Red Hat chose to develop it, I think, because developing a system in userspace is much easier, faster and flexible as opposed to landing something in the Linux kernel (similar to, say, how Google has decided to make build QUIC in userspace). I'm not here to advocate for or against it, because I have no interest in using it, but you might want to get some basic information first before commenting on it.

[@dixiannehawks1610]

ugh to the person who decided to focus on him instead of his laptop screen at 47:00

[@bertnijhof5413]

That high memory usage of ZFS is one of the urban myths. I used ZFS on an 2008 HP dc5850 with 8 GB of DDR2 and I was very happy with it. I allowed the memory caches to be resize dynamically between 0.5 and 1.5 GB dependent on other memory requirements. All datapools and caches were LZ4 compressed. Hit rates for disk io on the memory cache were 90%. I used BTRFS with LZO compression on a 32-bits system, a Pentium 4 HT at 3.0 GHz with two striped 40 GB IDE disks. Peppermint booted in 45 seconds and that is the fastest boot I ever saw on a Pentium. ZFS does not really support 32 bits anymore. I like both systems and the future developments will determine, which one I finally prefer.

[@unfa00]

If you want to have some fun, here's a bash command, that'll create a raid10 Btrfs filesystem across all USB drives connected to your system. mkfs.btrfs -f -m raid10 -d raid10 $(for i in $(lsblk -S | grep usb | cut -d' ' -f1); do echo "/dev/$i"; done) You can plug in as many USB flash memory sticks as you want and run this, it'll make a single filesystem across all of them. The SCSI disks are omitted. Try at your own risk. I'm having some fun with this right now to familiarize myself with Btrfs, as I've been only using ZFS so far.

[@galtbarber2640]

FYI zfs only uses lots of ram if you need block-level data de-duplication, which then creates a massive hash in memory for every block in the system. If you have a big storage cluster, that is a lot of ram. But if you do NOT use the feature, then its ram usage is not high.

[@unfa00]

I've deployed a few Linux machines on ZFS and have been using it for system root partitions and data for a few years now, but only recently I've learned that it's license is Linux-unfriendly. So I'm researching Btrfs, and I think I might actually migrate to it.

[@SNK1995]

Did you migrate?

[@unfa00]

@@SNK1995 Yeah. I've been on Btrfs for quite some time now. I love it :) And it doesn't need to be loaded as a module like ZFS, which is way better in case I need to do some troubleshooting.

[@eritert]

It is 2022 and I’ve just tried out btrfs for the first time, and he’s right, it is awesome (now).

[@askhowiknow5527]

If you don’t pronounce it ButterFace, you’re a monster. Also any journalling file system is good enough for me, really

[@kuhluhOG]

I pronounce it butterFuss (because Fuss sounds like Fass which is german for barrel)

[@justsomeguy7926]

I love all the questions that were asked. Most of which I would want to ask myself, but some that I didn't even consider. I especially liked the question on fault tolerance with btrfs-raid1 when you have more than one disk. I was a bit unsure when he first brought up the slide. I was thinking is it similar to a traditional RAID5 where you have parity? Or is it just duplicating the data across all disks randomly? I'm still not totally clear on this, but it's interesting. Something I can now look at.

[@LinuxIsNotAnOperatingSystem]

Man, I wish I could understand them. I could not pickup a single question from the crowd :/

[@alex.prodigy]

the RAM requirements for ZFS is exaggerated as always, if you have petabytes of storage then sure ... you might need 16GB or 32GB of RAM but lets face it , most small and medium sized companies are using volumes in the sizes of TB and you can easily get a way with like 2GB RAM on a ~10TB sized volume.

[@cepi24]

Is there any good reason why his terminal is not shown until around 47:00?

[@LinuxIsNotAnOperatingSystem]

and right on the commands part, I missed that.. :T

[@navneetpkl]

In the slide titled "raid1" != RAID1 (around 21st min), I see 2.5TB / 3 = 1.25TB usage. That seems absurd per my maths classes, it should either be 2.5TB / 3 = 833.33GB usage or 2.5TB / 2 = 1.25TB usage. Please advise which one is correct.

[@obvious_humor]

Also it says 500MB, lol

[@sondreavik5337]

1TB 1TB 500MB 1tb+1tb+500mb= 2.5tb, biggest drive 1tb, 1 drive fault tolerance = 2.5tb-1tb(biggest_drive)=1.5TB, not sure why it says 1.25TB it doesn't make sense to me at least, lol.

[@LinuxIsNotAnOperatingSystem]

It probably is 2.5TB / 2 = 1.25TB. Because it is a RAID 1 configuration, so you are using double the space for the same data.

[@MetroidChild]

Python isn't that bad (once you remove safety checks you don't need and when it's been compiled to .pyc), however I think the real use case is to use it as an interface or terminal for actually executing rust code.

[@Michael-wq3vb]

Pp

[@arlobubble3748]

ZFS ftw

[@katrinabryce]

You don't *need* 8GB of RAM for ZFS. My current daily driver is 4 x 10TB Iron Wolfs in Raid Z1. With 1GB RAM, it worked, but performance was a bit rubbish, about 60MB/s. With 6GB or RAM, it gives me about 230MB/s which is a lot more than my gigabit NIC can handle so absolutely fine for what I use it for.

[@Airbag888]

That's kind of a lot of ram for a file system... I'd be annoyed to have to reserve 6GB on my home server to not bottleneck my drives. That said I do not have 4x10TB and I wanted ssds mostly. Which I don't yet have. Oh boy. Anyway can you comment on the shortcomings of btrfs and how ZFS is able to suffer less?

[@jackchang3346]

Why not ZFS？

[@unfa00]

One reason is - because ZFS's license prohibits distributing it with the Linux kernel, which brings an overhead of using kernel modules instead. I am actually writing this on a Linux machine booting of off ZFS, but I'm considering switching to Btrfs at this point.

[@nacs]

In the video: 8:01

[@mikeg9b]

Why are you demonstrating Btrfs on a Windows computer? 47:01

[@anoopadv]

Always looked to me like pulseaudio and systems. Both of them made life harder. Now this one. It is always safer to have a fixed partition and say a daily backup. Simple enough. From hearing all this it is like having partition images on a disk. Compressed and expanded every read and write.

[@TitelSinistrel]

well...was completely open to btrfs and tested it a couple of times before, but after watching this talk and the logic behind it, I can say I'm now keeping my hands off it if this is the logic running in the background. Math doesn't add up, the way they treat the blocks/stripes is all wrong, it'll never be truly "stable".It's faulty from conception!

[@karakujacube]

excuse me, I do not have much experience with raid and filesystems (just a student), but where exactly is there a fault? The data seems always redundand in this "raid" version so why is it not stable? Could you please explain the flaw you mean?

[@gareginasatryan6761]

Biased against “the man”. In other words mainstream companies like MS and Apple. For example he complains that Apple “swindled” their customers by not backporting a new filsystem to an older OS. lolwut? Did Red Hat backport btrfs for older versions? He also compains that APFS wasn’t bootable. Well yes, it was still in development. The same way you couldn’t have zfs or btrfs on root for YEARS. Filesystems need extensive testing. It should be foolish for a company to hold off release until it’s feature complete, because they want as much public exposure as possible.

[@selforganisation]

I don't know the details, but if there was an OS update with no forewarning, then it wt wasn't an okay move. There are a number of ways to do this right, the simplest is to support the old FS and show an option to use the new one. Possibly with an in-place conversion tool, maybe integrated in the update process. Or a simple forewarning would do. Surely a user should expect an update to not make the hard drive inaccessible.

[@kc9scott]

@@selforganisation I have a Mac that I updated. I initially tried updating it to 10.14, which included in-place filesystem conversion of the (traditional/spinning) boot drive to APFS. There was (still is, AFAIK) a bug in 10.14, where a process ("iconservicesagent") will go crazy with infinite memory usage under certain conditions. I then had to take it to a shop for them to re-image the drive with 10.13, which can use HFS+ on this type of boot disk. This is not my main system, so I have not used it extensively since the update, but it seems to be working OK now. It was not a "push" update; you very much needed to decide when you wanted to do it, and there was a clear warning that you really should make a backup of your data before installing the update. I duly made my backup, and did in fact need it.

[@elimgarak3597]

Apple is garbage, tho

[@okinawadreaming]

lol, hes a linux user, what did you expect?

[@eritert]

ms and apple have literally swindled their customers, proven in court. After so many years it’s easy to just see them for what they are.

[@ermonnezza74]

BTW showing off your demo from inside a Windoze virtual box does not make it very appealing

[@ermonnezza74]

It is OK for a filesystem to be boring

[@liquidminds]

pics people upload on facebook must be horrible, if even their FS got STDs xD

[@TheChadWork2001]

These BTRFS presentations talking up BTRFS are a dime a dozen, and essentially useless. I appreciated it but where are all the ones showing how to use it? I've only found one and it had no explanations. We don't need to know why it's great. We need to know how to configure it in easy terms. What we need is explanations on how to partition with it and and use it. Demonstrations on how to set it up on computers, without having to dig for and read through a bunch of programmer complex jabber. Something common people can all understand. Even OpenSuse doesn't really explain BTRFS and acts like we all know how to use it as well as fdisk and regular partitioning.

[@Netist_]

You could always try reading the extremely simple documentation on the btrfs wiki.

[@hanro50]

Fedora has an option to outoconfigure an disk for you. I've been having a far smoother performence as of late thanks to it

[@Tom-kt8lu]

TheChadWork2001 Tell me which one!

[@shady4tv]

I hate to sound like an elitist snob but RTFM Seriously this is in ACTIVE development. I would not say this is very production ready but is gaining improvements on each kernel iteration. if you want something more documented and built out you should use zfs. btfs is still in the stage of figuring out what it is really doing and making core changes on each release.... Really the syntax of this is no worse than LVM. If you can get a hang of LVM you should be able to pick up btrfs pretty fast.

[@JivanPal]

@@shady4tv, uh, what? The spec has been stable since the end of 2013...

[@jeffbenzos1017]

you gave almost no info, wasting my time with what companies use it is far less relevant than fundamental concepts, implementation, frequency of code updates. quality of code, benchmarks across devices, etc . fucking hell. do you just like giving talks and dont really ccare if they are good?

[@--..__]

he put his entire presentation in dark mode... cringe

[@accelerat0r747]

clearly you've never heard of ZFS

[@nacs]

Clearly you didn't bother watching the video: 8:01