btrfs

[GReaper]

We've talked about ZFS before, anyone want to share their btrfs experiences?

I'm planning on using a single small SSD for / with ext4 (backed up to btrfs incase of hardware failure), my existing 2x1TB for /home with ext4 (I've had them a while, they're stable), and 2x3TB + some spare random drive with btrfs RAID1 for backups, downloads, and various things I'm not too bothered about losing.

Has anyone else here tried it?

 

[TdC]

ooh sekseh! I know of it, but since I use ZFS for just about everything I have not yet looked at it. Have you tried anything yet?

 

[caLLous]

I'm balls-deep in ZFS now otherwise I'd be very interested in it. I think it's getting more and more useable (I think it only just supports RAID5/6 in a stable fashion?) as time goes by and it definitely has some features that might put it a nose in front of ZFS one day like file-level cloning and the ability to expand arrays (and change the actual config of the array) on the fly.

Hah, I had visions of using 4 drives in October '13 with plans to expand to 8 at some point in 2014. I now have 22. :\

 

[GReaper]

Got the SSD in for / and EFI booting. Started off with an old 750GB drive, transferred some data, put in an old 400GB drive, rebalanced it to RAID1, currently transferring more data at the moment. I'll add in another 750GB later.

Hopefully the 2x3TB drives will arrive soon, I'll add those to the array and dump the two smallest drives (400GB and one of the 750GB). I'm using a Node 304 for the case - so there's only 6 drive bays, which is a limit I'm happy with.

I think it's getting more and more useable (I think it only just supports RAID5/6 in a stable fashion?)

RAID1 is mostly stable, RAID5/6 not so much - I think scrub/recovery for 5/6 has only been added in the past few months.

 

[MYstIC G]

Everything I've heard about it is bad. Turns up on the Linux action show every once in a while and never for a good reason. Personally I'd avoid it for anything that requires storing of value. Try it on an OS drive where you can rebuild without loss imho.

 

[MYstIC G]

https://rudd-o.com/linux-and-free-software/ways-in-which-zfs-is-better-than-btrfs

 

[caLLous]

Everything I've heard about it is bad.

That's a bit strong to put it mildly.

And that article must be ancient, some of the stuff it talks about just is not true any more. This one is from January last year and btrfs looks much more promising in it. I wouldn't put it in a production environment either but @GReaper said it was stuff he wasn't "too bothered about losing" so what's the harm?

 

[MYstIC G]

That's a bit strong to put it mildly.

No, it's a written representation of what I've heard about the file system.

This one is from January last year and btrfs looks much more promising in it.

Linux Action Show from July 2014 so some 6 months on from that article? http://www.jupiterbroadcasting.com/61572/preventing-a-btrfs-nightmare-las-320/

I especially like the links there to the reference material for things like: https://btrfs.wiki.kernel.org/index.php/FAQ#Why_is_free_space_so_complicated.3F

You know, where it says it's rubbish but you'll probably never get a reliable read on how much data is available for use?

I wouldn't put it in a production environment either but @GReaper said it was stuff he wasn't "too bothered about losing" so what's the harm?

Guess you missed the bit in bold:

Personally I'd avoid it for anything that requires storing of value. Try it on an OS drive where you can rebuild without loss imho.

Also the bold underlined bit means "in my humble opinion" because I guess you've forgotten what that means as well?

 

[MYstIC G]

Oh and sorry I forgot the "Gotchas" page https://btrfs.wiki.kernel.org/index.php/Gotchas every good file system should have one of those...

 

[caLLous]

Ok to all of that but we're clearly talking about testing here and they have beta/experimental warnings every time you do anything in btrfs to remind you of that.

There's lots wrong with btrfs (and it's positive that they know how much is wrong with it) but there's more right with it, when they get all of the bugs ironed out (and yes I know there's lots to fix). It IS going to be a superb filesystem. It's never going to make ZFS obsolete or anything but a proper native COW solution for Linux will make a real difference.

So don't put something like "everything I've heard about it is bad" because that's patently not true unless you've just been reading some naysayer or something. I've heard bad stuff too but also lots of good.

 

[MYstIC G]

Except it is true, so why are you attempting to make out that you know what I've heard better than I do?

 

[GReaper]

First impressions... not too bad.

I've added the 2x3TB drives, it was easy to add them to the btrfs array, and easy to remove the old devices. So much better than having to transfer data manually.

One thing I'm finding annoying is having to specify a mount point for the majority of btrfs commands. As I'm not using it as my root filesystem - it's frustrating having to specify the full path to a btrfs mounted directory.

Subvolume management isn't as great as it could be either, it feels a bit unpolished. It's also missing the one future planned feature which I'd love - different RAID levels per subvolume. If I could have RAID5 for common junk and RAID6 for things I care about - it would be a much better use of the storage I've got.

Overall it still isn't ready for mainstream usage. It's worthwhile testing on a server you can afford to lose some data, or on a server with good backups. I just hope I'll be saying that in a years time!