/tech/ - Technology★

>>889499 (OP)

what's wrong with ext?

>>889504

slow compared to xfs

Overall, XFS is better than EXT4... other than not being able to shrink volumes.

>>889546

>The Chad ReiserFS

>kills his wife and doesn't give a fuck.etc

>>889536

this.

ext is just spaghet torvalds code

it also fucks up on rolling distros (like getting errors after update) and fucks up again when kernel panics itself (being part of the kernel itself) while xfs doesn't do bullcrap.

>Best filesystem

>HAMMER2

Pick both.

How can it be the best filesystem if it can't even run windows

>>889592

>Has lost data at any point in its development

>best filesystem

>>889655

<Uses something marked in the documentation as unstable

<Encounters bug that makes him lose data

<wtf I hate btrfs now

Anything moderately complex is going to have its fair share of bugs. Just because some projects try to shove some under the rug doesn't mean that the amount of bugs in the past impacts the current state of the filesystem.

>>889499 (OP)

Honestly, I don't know much about filesystems or why one would be good over another. I've just been using ext4 for about 14 years or so and I never bothered to think about it. What would be a good primer on this?

>>889549

>it also fucks up on rolling distros (like getting errors after update) and fucks up again when kernel panics itself (being part of the kernel itself) while xfs doesn't do

Blame your shitty meme distro for that. Ext4 is as solid as a rock.

>>889546

Would it be possible to write to him and ask him to write code from prison?

>>889866

I've been using FFS/FFS2 (on OpenBSD) for same time, and I like not thinking much about the filesystem.

>>889536

ehh, I never have done benchmarks, so I didn't know how it was. Isn't xfs used on bsd's exclusively though?

>>889864

The whole filesystem was marked as unstable anon. I will give my argument as a comparison for (you).

Lets look at the goal and purpose of a filesystem. Its primary function is to store data, I don't anyone will disagree. Now lets look at something else, lets say a cryptography library or algorithm. The sole purpose of one of these is to keep data encrypted and secure, again I don't think anyone will disagree. Now if at any point in its lifetime the cryptography algorithm performs the opposite function and leaks your data, anyone logical and sane would stop using that library. Unless of course there were no other options. In which case you would need to evaluate if this is something you could recover from or if it would be better to start from scratch. Luckily this doesn't apply in this case. Lets go back to our filesytem example. We already know that the purpose of the filesystem is to store data, so we can continue to extrapolate that the opposite is to lose data. If at any point in its lifetime of being presented to anyone besides the developers it does the opposite of what its supposed to and loses data, anyone logical and sane would stop using it especially considering there are options available where that have never done this. Of course linux users have a long history of cuckoldry when it comes to this and tend to prefer the worse is better approach[1].

[1] http://minnie.tuhs.org/pipermail/tuhs/2017-May/009935.html

>>889542

Well, shrinking isn't safe on ext4 either.

>>889879

I've been using btrfs on my laptop and never experienced any data loss. Numerous times I've had the laptop forcefully shutdown without unmounting the filesystem. This includes times when it is actively writing to the disk.

>>889875

> Isn't xfs used on bsd's exclusively though?

I think it's the opposite. Wikipedo only shows experimental support for FreeBSD.

>>889884

ahh. Maybe that's why I just stick to ext.

>>889875

XFS is a remnant of SGI's IRIX. That's how fucking old it is. In Moronix benchmark tests, it is faster than Ext4, but it's a dinosaur. I prefer being able to grow and shrink my partitions and not having to worry about defragmenting my hard drives. This is why I would go out of my way to select EXT4 if XFS was default.

>>889499 (OP)

My biggest problem with XFS so far has been the number of programs (including GCC) that cannot cope with inode64. That is a major pain in the ass, especially when running gaymes in Wine.

>>889921

XFS still has fragmentation issues? Weird. I thought that was solved with journaling et al.

ext3 is fine

FAT is fine

just buy moar disk

>>889983

FAT can't handle really large files.

>>890041

The reference implementation can't. The reason FAT is the choice FS in embedded applications is because it's so barebones that there's little overhead and OEMs can just slap higher-level extensions on top of it.

>>889580

>implying NTFS is good

>implying FAT32 isn't obsolete on a HDD/SSD

>>889575

I don't have any choice but to accept.

>>889980

There is fragmentation, but not really an issue.

>not f2fs

You do have an NVMe SSD in CURRENTYEAR, right?

>>890555

I've been curious about using f2fs on my SSD. Is it natively supported in most distros?

>>890566

Yes.

>>890141

I thought he got life, but he actually plead down to second degree murder and got 15 years to life. So they could keep him in prison indefinitely, but he's eligible for parole in 2021.

I think he should have gotten a medal, not a prison sentence. Strangling a roastie is a public service.

>>890566

https://www.richud.com/wiki/Ubuntu_1504_Vivid_Install_To_F2FS_Filesystem

There might be a better way of doing this but I can't find one

>>889867

>solid as rock

>not lurk enough to read the mailing lists about the issues ext faced

tldr: torvalds: whoops, forgot to turn of blabla on the kernel, anyone who updated should reinstall before it's too late

>expects people to use enterprise distros when most pop distros are just enterprise testbed meant to squash bugs before it gets into the market

Don't even mention your distro of choice which is presumably Fedora

Any distro still defaults write caching on external drives and it's a wontfix kernel side problem. It's not just the fs but the kernel itself is shitting it up and even systemd helped worsen the problem.

>>889542

You can't safely or sanely shrink encrypted fs

>not FAT16

>>890566

Yep. GRUB doesn't like it though, so make your ESP /boot.

>>890607

> Strangling a roastie is a public service.

Beta virgin detected.

>>890731

Redhat/Cent/Fedora have switched to XFS. Why would I be pro EXT4 and use Fedora? Speaking of mailing lists, why don't you read some of Redhat's so you can see people shitting all over their decision to use XFS as default. It's nowhere near as reliable as EXT4. The only reason they (people in general) pick XFS over EXT4 is because it is faster. It certainly isn't for reliability.

>Any distro still defaults write caching on external drives and it's a wontfix kernel side problem.

If you unmount the drive before you yank the USB cord, its not going to be an issue. This isn't even a Linux or EXT4 specific problem. You can have data corruption on other file systems in Windows and OSX when not properly unmounting the volume.

>>889499 (OP)

Can confirm. I use XFS on everything. It's insanely fast on flash memory compared to EXT4. I get quicker startup times, faster file transfer on USB3.0 and SATA2/3, and applications just load a lot faster. Even on my laptop running Fedora KDE, Firefox will take about 5 seconds to open on EXT4. On XFS it's like 1 second.

Absolutely the best file system.

>>889504

>what's wrong with ext?

It's not btrfs, zfs, or refs.

>>892133

Aka bloated shit that include gzip, lvm, mdraid and everything into a big fat turd. Once XFS has CoW, it'll be perfect.

>>892136

>Once XFS has CoW, it'll be perfect

Copy on Write is a basic requirement, anything without it is fucking retarded.

>LILO can't be installed to VBR if boot volume is XFS

What the fuck is with this XFS bullshit. ZFS is the way and the light.

https://en.wikipedia.org/wiki/ReFS

When will ReFS become the default filesystem for Windows?

>>891088

>GRUB doesn't support f2fs

Jesus you're not kidding.

There's still patches being posted on the mailing list.

https://lists.gnu.org/archive/cgi-bin/namazu.cgi?query=f2fs&submit=Search!&idxname=grub-devel&max=50&result=normal&sort=date%3Alate

>>892864

I've run f2fs on my boot ssd for 6months on gentoo just fine???

>>892816

>ZFS is the way and the light.

but muh gnu-incompatible license

>>893127

just wait for oracle to relicense it

>>893130

They already did, newer versions have been proprietary for a few years now

>>893130

What everyone is using is actually "OpenZFS", because Oracle decided to do the things that Oracle does.

>>892839

Soon, hopefully, CoW is an indispensable thing, once you experience it.

>>889592

what i love most about btrfs are the bootable snapshots. as long as you take a snapshot before updating your system, there is no way you'll ever deal with a catastrophic system failure, short of a hardware failure.

>>889879

been using btrfs for 5 years on arch. never had a single problem.

>>889881

same here, numerous shutdowns without unmounting due to power failure over a period of years and i use dm-crypt on top of it and never had a data corruption problem. btrfs is simply the best there is and will only get better.

Which filesystem with CoW gives the best experience (as in stability and out-of-the-boxiness): XFS or btrfs?

>>893567

>i use dm-crypt on top of it

i should clarify that i meant i also use dm-crypt, but if you use whole disk encryption, you layer btrfs on top of dm-crypt, not the other way around

Looks like reflink is finally stable in 4.16. Pretty nice to see all the love XFS is getting recently.

>>893577

XFS, obviously. It's a lot stabler because it's

1) Made by SGI, who was a pretty good company

2) A lot older/tested (default on RedHat)

3) Less systemdesque in responsibility/functionality (as I said, btrfs like ZFS tries to replace stuff like LVM and mdraid without any reasons other than NIH).

>>893603

> default on RedHat

If even RedHat, which suffers a lot from NIH, dropped btrfs I don't see how the project can still have a future

>>893567

I was also using btrfs on top of dm-crypt too.

>>893605

FAT16. That's cool but I have a bit more respect for FAT64... err I mean exFAT.

Why doesn't ZFS get more love here?

>>894208

Personally for me it's a combination of its license and the fact that you need to dedicated a large amount of memory to it. Allocating the recommend amount of RAM for all my current harddrives in my computer would use up all of the RAM I have in it.

>>894208

I use ZFS for my NAS. It's really fucking nice.

ZFS does best when you build a machine dedicated to having it though, which typically isn't a general use machine.

No one talks about it here because:

-Years ago, there were annoying faggots that treated ZFS as a holy gospel that could do no wrong and you are a filthy fucking ext4 sinner,

-This place is filled with reactionary faggots, who will go to fantastic lengths to make mountains out of molehills about it's shortcomings because of the previous point.

Grub now has support for f2fs

https://www.phoronix.com/scan.php?page=news_item&px=GRUB-Now-Supports-F2FS