Is btrfs ready for / partitions?

[Starks]

I have a separate /home that I intend to keep as ext4, but I'm curious if all the stability and repair issues have been resolved for btrfs.

[davidmohammed]

no - stay well clear for btrfs for / - its known to take forever to boot from... some-sort of incompatibility with ureadahead or something like that.

If you must use it in ubuntu - use it from /home and keep to ext4 for /

[earthpigg]

http://arstechnica.com/science/news/...re-choices.ars

some dudes working on a $40,000 cluster say no.

they used ext4 for /, and (i think?) btrfs for /home - opposite of what you intend.

im sure they are also doing very frequent backups, including of /home, to mitigate risk.

[del_diablo]

Btrfs is ready for /, but it requires a ext2/3 /boot partition <3

[Starks]

Leading with /home is beyond stupid.

Your data is more valuable than your settings.

Also, I'm rather surprised that the /boot issue is still present. It should be resolved for Natty.

[jerenept]

http://arstechnica.com/science/news/...re-choices.ars

some dudes working on a $40,000 cluster say no.

they used ext4 for /, and (i think?) btrfs for /home - opposite of what you intend.

im sure they are also doing very frequent backups, including of /home, to mitigate risk.

ZFS > BTRFS right now imho.
This may change in the future though.

[earthpigg]

Leading with /home is beyond stupid.

Your data is more valuable than your settings.

Also, I'm rather surprised that the /boot issue is still present. It should be resolved for Natty.

i suppose that, for the slaves in a cluster, the configuration of / essentially becomes the vital data.

with /home as just a very large extension of the ram, intended to store things only temporarily until the results of the calculations are determined and then returned to the master. is that a correct understanding?

could explain why they are OK with btrfs as /home on the slaves, if its performance is significantly better and assuming bad mojo resulting from btrfs is detected so that the resulting conclusions from that subset of calculations can be discarded.

if that happens 3% of the time but makes the calculations 10% faster, it is a reasonable economic trade on the margin (the goal was a $40,000 cluster using off-the-shelf components that performed comparably to a $400,000 IBM). meanwhile, the custom software that runs the slaves is safe-and-sound from btrfs shenanigans in ext4.

[Starks]

ZFS is amazing filesystem if you overlook the fact that Oracle's license makes it difficult to incorporate into Linux.

[jerenept]

ZFS is amazing filesystem if you overlook the fact that Oracle's license makes it difficult to incorporate into Linux.

True. I suppose btrfs is trying to combat that?
If I need the features of ZFS, though, I would just run FreeBSD.

[BlacqWolf]

I have a separate /home that I intend to keep as ext4, but I'm curious if all the stability and repair issues have been resolved for btrfs.

I have previously had a good experience with a speed boost with BTRFS on 10.10 when I ran Ubuntu from a flash drive. But with my experience with it running on 11.04 on a virtual HD, it just doesn't boot, whereas XFS, JFS, and ext all work fine.