New XFS Programs Update Supports New XFS On-Disk Format

[phoronix]

Phoronix: New XFS Programs Update Supports New XFS On-Disk Format

After a year of development, xfsprogs 3.2.0 has been released as the latest version of the user-space program and other components for the XFS file-system. The big addition to xfsprogs 3.2.0 is supporting a new on-disk format...

http://www.phoronix.com/vr.php?view=MTY5MTY

[Paul-L]

As much as I like XFS, I now rarely use it for it's seemingly slow development, since ext4 has been improving and improving and being more and more both fast andreliable. Anyway it's nice to hear that improvements are still going after all.

[shmerl]

Are older deployments of XFS automatically upgraded to the new version when new kernel arrives, or it requires manual upgrade or not even possible?

[microchip8]

As much as I like XFS, I now rarely use it for it's seemingly slow development, since ext4 has been improving and improving and being more and more both fast andreliable. Anyway it's nice to hear that improvements are still going after all.

Bullshit. Subscribe to the XFS mailing list and see for yourself how much it gets developed. Let me give you a hint: daily arrive at least 10 patches by various people

EXT4 is not really that fast anymore. It's been getting slower and slower with each new kernel release. Sure, in some workloads it's faster than XFS, but in majority it's equally as fast. Also, XFS is targeted at enterprise audience, not really crappy pee-cee stuff. In all may years I have yet to see XFS eating up my data

People live under the illusion that EXT4 is somehow preventing all these things, while in reality EXT4 devs got a long way in making sure crappy apps that don't properly use fsync to commit data behave "safely" while on EXT4. XFS doesn't really do that, and IMO, it's the job of the app developer to make sure user data is properly committed to disk, not the other way around (FS making sure crappy dev's program commits data safely to disk at proper intervals and operations)

[Pseus]

As much as I like XFS, I now rarely use it for it's seemingly slow development, since ext4 has been improving and improving and being more and more both fast andreliable. Anyway it's nice to hear that improvements are still going after all.

I am not so sure development has been slow. Check the mailing list.

[microchip8]

Are older deployments of XFS automatically upgraded to the new version when new kernel arrives, or it requires manual upgrade or not even possible?

No and no. you have to reformat the partition with the xfsutils that support the v5 format, ie by passing -m crc=1 to mkfs.xfs

[microchip8]

As much as I like XFS, I now rarely use it for it's seemingly slow development, since ext4 has been improving and improving and being more and more both fast andreliable. Anyway it's nice to hear that improvements are still going after all.

Not true on all counts.

1. XFS is moderately to heavily developed. Check the mailing list. Hint: at least 10 patches are posted daily
2. EXT4 has been getting slower and slower with each kernel release while XFS has stayed at its performance level (maybe too getting slightly slower but not to the same amount as EXT4)
3. XFS encourages proper programming wrt dealing with data safety instead of encouraging bad programming like EXT4 does by working around how programs that don't care much for user data store it to disk

Also XFS is really targeted at enterprise levels where redundancy is almost always present. That said, I never had problems with XFS on PC hardware

[birdie]

More importantly: does it shrink?

It's my most significant concern before I ever start using it again.

That said, I never had problems with XFS on PC hardware

Once I lost over 150GB of data on RHEL 5 when I was using it (six years ago). The data was still on the disk but all its metadata magically disappeared without any interference on my part. YMMV, huh?

[Paul-L]

Not true on all counts.

1. XFS is moderately to heavily developed. Check the mailing list. Hint: at least 10 patches are posted daily
2. EXT4 has been getting slower and slower with each kernel release while XFS has stayed at its performance level (maybe too getting slightly slower but not to the same amount as EXT4)
3. XFS encourages proper programming wrt dealing with data safety instead of encouraging bad programming like EXT4 does by working around how programs that don't care much for user data store it to disk

Also XFS is really targeted at enterprise levels where redundancy is almost always present. That said, I never had problems with XFS on PC hardware

1. What I meant, slowly developed in comparison of ext4.
2. Check the benchmarks on I/O in this very site. It's not the same.
3. As a home user and hobbyst on programming, ext4 has become faster (atleast for me, obviously) dealing with my personal problems; I know that XFS is targeted to enterprise environments and that's why I like it so much, redundancy.

I meant no offense of course, just that in my point of view, patches submitted that are not merged mean nothing; and for some reason ext4 has become faster on my volumes while XFS has become slower (with some performance improvements on some kernel releases). Again, this is my personal point of view, not to say that what you written is not true for your particular enviroment.

[microchip8]

More importantly: does it shrink?

It's my most significant concern before I ever start using it again.



Once I lost over 150GB of data on RHEL 5 when I was using it (six years ago). The data was still on the disk but all its metadata magically disappeared without any interference on my part. YMMV, huh?

No, it does not shrink yet at the moment. There are patches available for that (check XFS wiki) but they're not ready yet. AFAIK, no one is working on that currently.

About the loss of data, yes XFS had problems with that back in the past. It was due to bugs which have been fixed in mid 2007. In addition, XFS tended to perform really bad when lots of metadata was involved as the journal was getting hit with only metadata and couldn't process anything else. After a few failed tries, Dave was able to significantly redesign how XFS treats metadata (he basically borrowed the concept from EXT4 as it was proven to work and there was no need to reinvent the wheel)