No, a filesystem with send/receive support should be ubiquitous in backup usage.
Cold data in such a scenario should have zstd-19 applied, it's cheap to reheat.
Super slow and unfinished. Though it may have been the case of cache SSD being slow, but when I was copying data back to BTRFS the average speed was 17Mb/s.
It failed at watching an episode of a series.
I'm back to BTRFS + BCACHE for now.
Don't CoW filesystems suffer from awful free space fragmentation if used with files like VM images or torrents, or any large files which can have a bunch of random writes happening all throughout?
Nah, it's because you can only 80% of a disk normally then need to either eat a massive performance loss or "rebalance"/defragment.
Any CoW fs above 95% usage is simply unusable.
CoW fragmentation doesn't work like overwriting filesystems, any performace loss is already spent by the design.
Btrfs is closer to a regular filesystem and will use 1GB buckets to "rebalance" blocks and ZFS is a block database that uses caching to alleviate performance issues.
B-epsilon trees are allegedly outright fixing any performance issues with fragmentation and filesystems entirely.
>fell for the butterfs meme >accidentally erase a btrfs and an ext4 drive during a windows install >fully recovered the ext4 drive, no data lost >butterfs drive is lost is lost forever.
Frick this israeli glowBlack person scam
Selected the wrong drives when installing windows, an though the installer was like calamares where you could redo and abort the formatting of the drive without losing any data. Big mistake.
>kernel 6.7 is a time bomb upgrade to 6.8 now >replication is stable, erasure coding is not >compression is single-threaded so stick to zstd:1 or rebalance thread will be piss ass slow >ssd caching works very well >can decide filesystem options like foreground/background target or replication per folder, like make sure folder x will only live on ssds, disable replicas on temp download folders >is not block based and not fixed layout, so you can expand or remove drives, mixed drive sizes, no issues
its quite a good "throw a bunch of drives on the server" filesystem
what is the advantage over zfs
I've been using btrfs for I think half a decade. It's great
License I suppose
easier rebalancing, more suitable for poor people like me with hdd of various size and quality.
it can do automatic mutt raid
I was able to move all my data back to BTRFS by converting to RAID0 and then adding more drives and converting to RAID5
Yes, I use btrfs. No, I have no idea how it works/what is is. archinstall recommended it.
I'm on ext5 beta
just works
used years ago
broke my files
Why would I use anything but NTFS
Enjoy defragging your shit fragile FS
I'm waiting to have problems with ext4 to switch, but it's been taking a while.
I am too dumb to use btrfs.
Simple as.
I use BTRFS on my NAS with RAID 1 and I use ext4 on my single backup HDDs. Am I doing it right?
No, a filesystem with send/receive support should be ubiquitous in backup usage.
Cold data in such a scenario should have zstd-19 applied, it's cheap to reheat.
copy-on-write send and receive is the spaceman to pluming on multiple filesystems' cowboy.
ext4 is fast but can break easily with few bad sectors
btrfs is more reliable but slower, especially if you create many snapshot, docker can also take advantage of snapshots
My only issue with btrfs is not having a clear disk usage indication
btrfs/zfs handles I/O pressure much better despite being "slower" than e.g. xfs, benchmarks don't show that
>not using bcachefs
Bcachefs is not stable.
i use OpenSUS + btrfs btw
Super slow and unfinished. Though it may have been the case of cache SSD being slow, but when I was copying data back to BTRFS the average speed was 17Mb/s.
It failed at watching an episode of a series.
I'm back to BTRFS + BCACHE for now.
Made in my new Artix Install, no day-a-day benefits seen so far.
just werks.
ext4 on root and zfs on data drives.
i prefer not to use "totally stable bro" for my data
lvm snapshots do the job just as well
This shit bricked my system after a month of use.
why is there a filesystem called buttfrickers
Don't CoW filesystems suffer from awful free space fragmentation if used with files like VM images or torrents, or any large files which can have a bunch of random writes happening all throughout?
You're expected to manually disable CoW on directories containing those types of files.
I'm new to all this, but is this why people say that CoW are great, but require more effort put into their usage or they become a hindrance?
Nah, it's because you can only 80% of a disk normally then need to either eat a massive performance loss or "rebalance"/defragment.
Any CoW fs above 95% usage is simply unusable.
CoW fragmentation doesn't work like overwriting filesystems, any performace loss is already spent by the design.
Btrfs is closer to a regular filesystem and will use 1GB buckets to "rebalance" blocks and ZFS is a block database that uses caching to alleviate performance issues.
B-epsilon trees are allegedly outright fixing any performance issues with fragmentation and filesystems entirely.
>fell for the butterfs meme
>accidentally erase a btrfs and an ext4 drive during a windows install
>fully recovered the ext4 drive, no data lost
>butterfs drive is lost is lost forever.
Frick this israeli glowBlack person scam
>letting windows touch your hard drives directly
You fricked up.
erase a btrfs and an ext4 drive during a windows install
You said it yourself, lmao
Explain in detail.
Selected the wrong drives when installing windows, an though the installer was like calamares where you could redo and abort the formatting of the drive without losing any data. Big mistake.
This got nothing to do with btrfs or ext4, it's entirely Windows' fault.
why are you dual booting on your fricking server? which is where btrfs is used
I'm not gonna use beta tier at best software for my filesystem.
>b-but
I don't care, ZFS just works. Maybe your data is worthless, mine isn't.
Of course I'm using btrfs, with snapper, compression and deduplication
I don't want to get my files destroyed.
im using bcachefs on my server
>kernel 6.7 is a time bomb upgrade to 6.8 now
>replication is stable, erasure coding is not
>compression is single-threaded so stick to zstd:1 or rebalance thread will be piss ass slow
>ssd caching works very well
>can decide filesystem options like foreground/background target or replication per folder, like make sure folder x will only live on ssds, disable replicas on temp download folders
>is not block based and not fixed layout, so you can expand or remove drives, mixed drive sizes, no issues
its quite a good "throw a bunch of drives on the server" filesystem
What's the drive set up? I had terrible performance with 4x 18Tb.
yeh performance is ass without ssds to handle foreground writes and metadata.
but performance was really ass in 6.7 either way, 6.8 fixed a lot.
im running with okay performance with
2x 2tb intel p3600
8x mixed hdd, mostly 6tb
I'm using UFS, works fine.
>my os so fragile it constantly needs backup
>btrfs for /
>xfs for other drives
Am I doing it right?
Yes. That's the OpenSUSE default.
burden of proof is on you
why should i use btrfs over ext4 when it served me well for the past 10 years?
Why would I? If it works don't fix it
>NTFS