Bcachefs File-System Pull Request Submitted For Linux 6.5

[piorunz]

I think I understood this. But a SATA cable is unlikely to fail and you can easily find out if the cable is defect by switching both and see if the error happens on the other drive. If not it's not you cables and either your Motherboard (SATA slot? -> switch to unused slots) or the drive (you know S.M.A.R.T.?) is the reason.
You didn't seem sure about that. Also don't waste money for SATA cables, whomever builds a PC has one lying around.

Normally you use raid for high availability and a restart of that machine is what you want to avoid. Do you have a hot plug option for SATA in you BIOS? Without raid is pretty useless.

You may have other reasons to use raid but you should know it doesn't replace a backup.

Thanks for your unsolicited advice. And now let me use my raid setup as I want to, thanks

[timofonic]

You can view the lwn article in the provided link on the end of my message.

Forget about the BSD vs GPL nonsense, it seems trolling to me and a derail of the main topic. I was terribly bored anout BSD/Sun/Solaris nostalgia, OpenSolaris/Illumos "activists", GPL/BSD haters, Oracle criticism (I know who they are and ignore their stuff, I'm not so idiot) and such. I respect them and their stuff, but that evangelism and constantly repeating topics drives me away and I ignore it no matter what side comes from.

https://news.ycombinator.com/item?id=36366002

[Quackdoc]

Thanks for your unsolicited advice. And now let me use my raid setup as I want to, thanks

imagine posting something on a forum and when someone comments on it, this is the trash you reply with.

[ihatemichael]

Strongly recommended for VM-images and database workloads since they do not work well with copy-on-write. Beware especially of running copy-on-write-on-copy-on-write, i.e. btrfs VM on a btrfs host. Don't do it. Good praxis is to set 'chattr +C' on the folder where you keep your VM's, but keep in mind that the attribute is only set on new files created in the folder, not already existing files. For your old VM's, you can work around that by first setting 'chattr +C' on the folder, rename your old VM-image, example append '.backup' to their name or similar, and make a copy of it with it's original name. Confirm with 'lsattr' that they have the '+C' attribute. And keep in mind it's uppercase 'C', as lowercase means compression, not nodatacow.

Thanks for the advice, I will look into those things.

[ihatemichael]

I'd second what Brisse above had said.

If you can live without checksumming (that's what you'd get by switching to XFS or ext4 anyway), you can simply turn that feature off for selected files (e.g. disk images or database files) and you'll get the approximately the same performance as you'd get with XFS or ext4.

You still get some fragmentation in case you make snapshots (a feature that's not available on ext4), but if that is a problem, you can use `btrfs fi defrag` to fix it after snapshotting.

I think I can live without checksumming, I certainly can live without snapshotting.

I am not sure I can live without subvolumes, they really make things a lot easier.

I currently dual boot more than one distro on the same machine and subvolumes are easy to create without having to repartition the whole drive.

Subvolumes also make it possible to use the drive space in a more intelligent way, you don't need to estimate how much space each partition will use.

I can probably achieve some of those things with LVM but I am not sure it's worth it.

I certainly don't want to go back to adding 4 or 5 partitions per drive.

[Anux]

I can probably achieve some of those things with LVM but I am not sure it's worth it.

I use LVM to pass through disk space for my VM, it's the best performing method after having independent physical disks for each VM.

[darkbasic]

The ability to roll back to an arbitrary previous version of a generic file is remarkably useful.

With btrfs I work this around by doing cp --reflink=always /path/to/snapshot/file /path/to/file

[darkbasic]

Strongly recommended for VM-images and database workloads since they do not work well with copy-on-write.

What's the point of btrfs if you use nodatacow? Why can't it run vms and databases with decent performance like zfs does? zfs is cow as well.

[darkbasic]

You still get some fragmentation in case you make snapshots (a feature that's not available on ext4), but if that is a problem, you can use `btrfs fi defrag` to fix it after snapshotting.

No you can't because defragmenting with snapshots means that if your disk usage was 10 GB it will become 100 GB with just 10 snapshots.

[darkbasic]

I think I understood this. But a SATA cable is unlikely to fail

You would be surprised to know how many failures I had due to sata cables or controllers...