I think this is a political move disguised as technical move
oracle pays the developers of btrfs [0]
redhat hates the guts of oracle, since oracle released oracle linux, which is a clone of redhat enterprise (based on centos)
so, redhat wants to cripple btrfs and hurt oracle.
However, btrfs is my favorite FS, been using it on my home computer and backup drives for at least 6 years, before it was included in the kernel, love the subvolumes, snapshots, and compression; never had issues with it
.
I think this is a political move disguised as technical move
I think there are solid technical reasons to discourage Btrfs use, just to quote from the official wiki [0]:
> The parity RAID code has multiple serious data-loss bugs in it. It should not be used for anything other than testing purposes.
Now I don't know if this issue has been addressed already, or which kernels are affected, but the fact that there is a prominent warning on the wiki speaks for itself.
Personally, I'm a happy btrfs user deploying a mixed-disk-size array without parity, with the hope to add redundancy some time in the future. Currently, btrfs is the only FS allowing to mix disks of any size and to run an optimal configuration on top of them [1].
The particular bug that sparked that warning was fixed a while ago, but as a precaution against "btrfs ate my data" stories they've removed the ability to create btrfs-raid from the CLI tools (you can still use md RAID with btrfs but you lose most of the benefits of btrfs that way).
The design of btrfs allows for mismatched disks to be used in an array, and the btrfs RAID will keep the right level of redundancy while using the maximal amount of space, e.g. with a 4 TB, 2 TB, and 1 TB drive:
mdadm will give you 1 TB in RAID 1, or 1.5 TB in RAID 10 (constrained by the smallest drive).
btrfs will give you 3 TB in RAID 1 (constrained by the sum of the smallest drives).
btrfs also allows per-subvolume raid policies. So you could, for example, give users an "archive" subvolume in their home directory. You could then mark this as RAID 1 or RAID 5 (because you don't care so much about performance) while the main /home filesystem is RAID 10.
Unfortunately the RAID code is all horribly broken.
Being able to have non-symmetric disk topologies with redundancy. I believe that md raid does not support that, while btrfs multi-device does (which is what I think of as one of the really unique features of btrfs -- not even ZFS can handle the sort of disk topologies that btrfs can).
md-raid absolutely supports that and synology has for a long time. They call it "SHR". You simply do raid over disk partitions to enable disks of disparate sizes.
The reason ZFS doesn't support it, and absolutely 0 enterprise storage devices support this is because as the disks fill up, you sacrifice both performance and redundancy. Synology won't even support it on their high-end devices for this very reason. They'll only do it on their devices targeted at home use.
"In June 2012, Chris Mason left Oracle for Fusion-io, which he left a year later with Josef Bacik to join Facebook; while at both companies, Mason continued his work on Btrfs."
> However, btrfs is my favorite FS, been using it on my home computer and backup drives for at least 6 years, before it was included in the kernel, love the subvolumes, snapshots, and compression; never had issues with it .
Slightly off topic. I chose btrfs as my main filesystem recently on a system running Ubuntu/Xubuntu. I have done some research on backing up (with the advantage of snapshots) but it looks like there aren't (m)any graphical tools (this gets a little more confusing with /@ and /@home subvolumes on the same partition being treated separately for snapshots, AFAIK).
Do you manage it all from the command line and/or do you have any suggestions for graphical tools to do "as-is clones of entire partitions" (and also incremental backups) to local external drives (not over the network)? Or if you could point to any great documentation or blog posts on this topic, that'd be helpful too (I have read some bits of the btrfs wiki and the btrfs parts in the Arch wiki).
Currently I'm doing a plain rsync using Grsync, and not really taking advantage of btrfs features like snapshots.
The main reason I'm looking at avoiding the command line is to make it easier for others around me to use it.
performance during snapshot creation/deletes (presumably mostly the deletes) is one of the reasons I personally stopped using btrfs on my desktop. Now using ZFS root (with Ubuntu devel).
I had auto hourly snapshots and sometimes when it deleted one my entire system would hang for a few seconds and occasionally 10s of seconds.
Having said that I do suspect that might be partially related to also using ecryptfs on top, but still.
If anything, the political move would've been Oracle's sponsorship of btrfs in the first place. They want to push people to enterprise OS/storage systems, so they've told everyone "Oh yeah, btrfs is coming soon, it'll be great" ... and it isn't great. It sucks.
I've finally broken and installed ZOL (ZFS on Linux) after trying btrfs repeatedly over the last three years. ZFS is already a breath of fresh air and I've only been using it a couple of months. For whatever reason, btrfs came together as a messy hodge-podge, and it shows in bad performance for many use cases (e.g. "omg I forgot nodatacow"), buggy implementations, difficult user interfaces, kernel bugs, etc.
btrfs needs a reboot (I hear bcache? is trying). Meanwhile, everyone should stop getting hung up on the arcane licensing details and just use ZFS directly. It can't be distributed as part of the kernel, but that's why we have distributions, isn't it? They bundle all that crap together for us. There shouldn't even be the normal OSS infighting because this isn't a proprietary blob or something, it's just using a license that's GPL-incompatible.
The best thing Linus could do for the community at large would be to fork and start committing to ZOL, giving it a tacit endorsement.
> I think this is a political move disguised as technical move
You literally have no idea what you're talking about, and I doubt you've used btrfs seriously, or you wouldn't talk this shit. The fact it's been upvoted so heavily just shows what absolute technically-false nonsense will draw support at HN.
oracle pays the developers of btrfs [0]
redhat hates the guts of oracle, since oracle released oracle linux, which is a clone of redhat enterprise (based on centos)
so, redhat wants to cripple btrfs and hurt oracle.
However, btrfs is my favorite FS, been using it on my home computer and backup drives for at least 6 years, before it was included in the kernel, love the subvolumes, snapshots, and compression; never had issues with it .
[0] https://oss.oracle.com/~mason/
[Update] Chris mason no longer at Oracle since 2012