Fedora Server 22 Is Using The XFS File-System By Default

[caligula]

XFS loses to EXT4 in all phoronix benchmarks. Maybe not relevant in servers. I'd use ZFS instead on servers and EXT4 on desktop. Btrfs maybe when it's done.

[edgan]

The first link below goes into detail about various filesystems and data journaling, in the Filesystems section. It brings up that XFS can do ordered in certain cases, but doesn't guarantee it always will be.

The second link asks the same question of is XFS metadata only, the equivalent of write-back.

http://monolight.cc/2011/06/barriers-caches-filesystems/

http://serverfault.com/questions/680102/xfs-as-default-filesystem-in-the-enterprise

[SyXbiT]

Banks, stock exchanges, governments, etc.. etc.. etc.. all use RHEL.
Do you realize just how much money and testing had to happen before Red Hat decided to move RHEL to xfs (and by association, Fedora Server)?
A LOT. It's a huge decision. I don't think they took the decision lightly.

[rvdboom]

My personal experience is 8+ years using XFS on all my desktop, server and laptops, never lost data on any of them even during complete outage of the whole neibourghood.
It doesn't mean a lot statistically, of course.
I would guess that XFS is designed to scale very well with FS size. Probably very interesting nowadays for servers.

[pdffs]

XFS has certainly eaten all my data, and if there is FS corruption, the repair tools could make it significantly worse. This was all a number of years ago though, because I stopped using it after multiple such experiences.

[Awesomeness]

The only mistake was to develop the fucking stop-gap file system ext4 in the first place. It was only intended to be used for a year or so because (obviously?) btrfs 1.0 was to be released in 2008. Just like ext3 before, it just tacked more shit onto ext2.
Wasting resources on ext4 is among the reasons why to this day no credible Linux distribution trusts btrfs to handle /home

[RahulSundaram]

The only mistake was to develop the fucking stop-gap file system ext4 in the first place. It was only intended to be used for a year or so because (obviously?) btrfs 1.0 was to be released in 2008. Just like ext3 before, it just tacked more shit onto ext2.
Wasting resources on ext4 is among the reasons why to this day no credible Linux distribution trusts btrfs to handle /home

Would you consider SLES and openSUSE credible?

[Guest]

XFS loses to EXT4 in all phoronix benchmarks. Maybe not relevant in servers. I'd use ZFS instead on servers and EXT4 on desktop. Btrfs maybe when it's done.

Are you sure about this?

I looked into benchmarks in the past, and I could have sworn XFS came out on-top performance-wise over ext4 in majority of the benchmarks.

[nanonyme]

The only mistake was to develop the fucking stop-gap file system ext4 in the first place. It was only intended to be used for a year or so because (obviously?) btrfs 1.0 was to be released in 2008. Just like ext3 before, it just tacked more shit onto ext2.
Wasting resources on ext4 is among the reasons why to this day no credible Linux distribution trusts btrfs to handle /home

Well, kind of. Imo Ext3 was the (easy to say in retrospect) mistake. Ext4 and Ext2 both are very useful filesystems. Ext2 as a bare unjournaled one and Ext4 as an efficient storage. Ext3 isn't really better than Ext4 in any conceivable way. Btrfs is a good pick already but its CoW nature makes it a bad fit for DB's hence you most likely only run it for a file server, not a DB server. We're further and further down the path where you can't just pick one filesystem and use it as answer for everything

[xeekei]

Ext4 can't scale for shit. It is good for home use or small servers, but when you're talking hundreds of terabytes Ext4 is not even in the race. The only weakness XFS has is its inability to shrink, but you never/very rarely have to shrink a partition in enterprise. Hopefully XFS will get a boost in development now so we can have an awesome non-COW alternative in the future.