ZFS device removal is almost here:

[brutalizer]

Adam Leventhal (one of the architects of DTrace) has tweeted:

"teaser: OpenZFS device removal has just landed in our repository; looking forward to seeing it upstreamed!"

 

[FnordMan]

Adam Leventhal (one of the architects of DTrace) has tweeted:

"teaser: OpenZFS device removal has just landed in our repository; looking forward to seeing it upstreamed!"

Shiny, wonder when it'll show up in ZFS on Linux, not that I really need that feature. All they really need now is the ability to do the opposite (grow an array) and btrfs will be utterly pointless.

 

[arnbju]

It's about removing a vdev from a pool. Not shrinking the size of a vdev.

Slides here: http://open-zfs.org/w/images/b/b4/Device_Removal-Alex_Reece_&_Matt_Ahrens.pdf (demo on the last slide)

 

[SirMaster]

http://blog.delphix.com/alex/2015/01/15/openzfs-device-removal/

 

[cantalup]

Shiny, wonder when it'll show up in ZFS on Linux, not that I really need that feature. All they really need now is the ability to do the opposite (grow an array) and btrfs will be utterly pointless.

I put my faith on btrfs. growing rapidly as I know...
coexist betwen zfs(zol) and btrfs religion...

 

[drescherjm]

I get the impression that development on btrfs is slower due to less active developers. Also it does not appear that btrfs is tested anywhere near the level that zfsonlinux gets tested.

 

[cantalup]

I get the impression that development on btrfs is slower due to less active developers. Also it does not appear that btrfs is tested anywhere near the level that zfsonlinux gets tested.

it is pacing up .....

ZoL is slow too...check their history release level....

btrfs 0/1 is pretty solid in general . I am testing on my machine since centos7 support btrfs .

ZoL basically identical, not really gets tested throughly, only user (you, other, and I) are doing stress testing and report bugs.

I am Happy with ZoL and btrfs in general, just to know where I can push to the limit, not less/more.

 

[bexamous]

I thought vdev removal and defrag were the same problem?

 

[raiderj]

I thought vdev removal and defrag were the same problem?

I would think that you could remove a vdev without having to defrag/rebalance.... but it would make sense to do so. Rebalancing in general would be nice for when you add more vdevs to a pool too.

 

[arnbju]

I would think that you could remove a vdev without having to defrag/rebalance.... but it would make sense to do so. Rebalancing in general would be nice for when you add more vdevs to a pool too.

Rebalace would also help to enforce compression or dedup for data that was written before setting the option.

 

[Aesma]

Yeah to me it seems defrag has to be linked and that's what I'm wondering about.

 

[raiderj]

I think you could do a vdev removal *without* a defrag by just copying over data from the targeted vdev to another location. But, that's obviously non-optimal.

I wonder why rebalancing/defrag wouldn't come first though. That would be very helpful for vdev additions as well.