I'm migrating to new large drives and would lke to take advantage of a filesystem that has integrated data and metadata checksums. How well does ZFS on Linux really work? What is the current status of BTRFS? Which one would you trust your precious data to? I've been out of the loop lately... This would be a great meeting topic! "How to set up ZFS on Linux and/or BTRFS". Any takers?
Since I use zfs on a daily basis, I'd be glad to talk about it. How big are these drives? While you can use zfs as your boot FS, I think I'd stick with lvm/ext4 for / and / boot. I'd then carve out a big data space and use zfs for that. Are these disks going to raided? Thanks, Tim. On Tuesday, April 28, 2015, Chuck Anderson <cra@wpi.edu> wrote:
I'm migrating to new large drives and would lke to take advantage of a filesystem that has integrated data and metadata checksums. How well does ZFS on Linux really work? What is the current status of BTRFS? Which one would you trust your precious data to? I've been out of the loop lately... This would be a great meeting topic! "How to set up ZFS on Linux and/or BTRFS". Any takers? _______________________________________________ Wlug mailing list Wlug@mail.wlug.org http://mail.wlug.org/mailman/listinfo/wlug
-- I am leery of the allegiances of any politician who refers to their constituents as "consumers".
2 x 4TB, RAID 1, I prefer to boot from them (hmmm...although not booting from them is an excuse to buy an SSD for the boot drive...) I already use ZFS on FreeNAS, but I'm wary of using ZFS on Linux since it is a third-party addon, you must recompile it every time the kernel changes, etc. I never liked having to do that for the Nvidia driver. Losing your video display is one thing, losing access to your data is much worse... On Wed, Apr 29, 2015 at 06:19:19AM -0400, Tim Keller wrote:
Since I use zfs on a daily basis, I'd be glad to talk about it.
How big are these drives? While you can use zfs as your boot FS, I think I'd stick with lvm/ext4 for / and / boot. I'd then carve out a big data space and use zfs for that. Are these disks going to raided?
Thanks, Tim.
On Tuesday, April 28, 2015, Chuck Anderson <cra@wpi.edu> wrote:
I'm migrating to new large drives and would lke to take advantage of a filesystem that has integrated data and metadata checksums. How well does ZFS on Linux really work? What is the current status of BTRFS? Which one would you trust your precious data to? I've been out of the loop lately... This would be a great meeting topic! "How to set up ZFS on Linux and/or BTRFS". Any takers?
Chuck Anderson <cra@WPI.EDU> writes:
I'm migrating to new large drives and would lke to take advantage of a filesystem that has integrated data and metadata checksums. How well does ZFS on Linux really work? What is the current status of BTRFS? Which one would you trust your precious data to? I've been out of the loop lately... This would be a great meeting topic! "How to set up ZFS on Linux and/or BTRFS". Any takers?
I wouldn't trust my data to btrfs, especially if you're using anything other than the default mkfs and mount options. But don't take my word for it, have a look through the btrfs mailing list archives[1]. I have no experience with zfs on linux, but I've heard that data checksums have a signficant performance impact. I don't have any data on that, so you should probably do your own benchmarking to verify things perform well enough for your workload. If I were you, I'd use xfs. Cheers, Jeff [1] http://marc.info/?l=linux-btrfs
Hi Jeff, its nice to provoke a response out of you...when are you coming to a meeting? :-) On Wed, Apr 29, 2015 at 09:40:09AM -0400, Jeff Moyer wrote:
I wouldn't trust my data to btrfs, especially if you're using anything other than the default mkfs and mount options. But don't take my word for it, have a look through the btrfs mailing list archives[1].
I see what you mean. People are still asking questions "how do I fix this broken fs?" and being told to upgrade to linux kernel 4.0.
I have no experience with zfs on linux, but I've heard that data checksums have a signficant performance impact. I don't have any data on that, so you should probably do your own benchmarking to verify things perform well enough for your workload.
If I were you, I'd use xfs.
I don't like XFS because it can't be shrunk in place. Too many times I've needed to shrink one LV to make space for another LV.
Chuck Anderson <cra@WPI.EDU> writes:
Hi Jeff, its nice to provoke a response out of you...when are you coming to a meeting? :-)
Heh. Not sure. I had considered asking if the group would be interested in a talk on persistent memory.
If I were you, I'd use xfs.
I don't like XFS because it can't be shrunk in place. Too many times I've needed to shrink one LV to make space for another LV.
OK. You could use dm-thinp for this, but I understand an aversion to adding another layer. Cheers, Jeff
It seems to me that you want both resiliency (able to suffer a drive failure) and checksums of the files you have on there as well. Would it make more sense to use 'tripwire' with an ext4 filesystem? I know that rock solid filesystems are my choice. I've looked at using XFS and JFS in the past... but the downsides were just too much for my tastes. Esp since I do like to upgrade kernels and such. Being able to resize the filesystem (offline is fine) is a winner. Online resizing would be even better, but that's a much tougher nut to crack. Chuck> Hi Jeff, its nice to provoke a response out of you...when are you Chuck> coming to a meeting? :-) Chuck> On Wed, Apr 29, 2015 at 09:40:09AM -0400, Jeff Moyer wrote:
I wouldn't trust my data to btrfs, especially if you're using anything other than the default mkfs and mount options. But don't take my word for it, have a look through the btrfs mailing list archives[1].
Chuck> I see what you mean. People are still asking questions "how do I fix Chuck> this broken fs?" and being told to upgrade to linux kernel 4.0.
I have no experience with zfs on linux, but I've heard that data checksums have a signficant performance impact. I don't have any data on that, so you should probably do your own benchmarking to verify things perform well enough for your workload.
If I were you, I'd use xfs.
Chuck> I don't like XFS because it can't be shrunk in place. Too many times Chuck> I've needed to shrink one LV to make space for another LV. Chuck> _______________________________________________ Chuck> Wlug mailing list Chuck> Wlug@mail.wlug.org Chuck> http://mail.wlug.org/mailman/listinfo/wlug
participants (4)
-
Chuck Anderson -
Jeff Moyer -
John Stoffel -
Tim Keller