> From: Discuss [mailto:discuss-bounces+blu=nedharvey@blu.org] On
> Behalf Of Tom Metro
>
> (Is there any other solution outside of a NetApp file
> or BtrFS compare in this area? Maybe with vast quantities of cheap
> storage, the space inefficiency of snapshots is less of a concern.)
Yeah, lots. MS uses volume shadow services. And all the big guys (isilon etc)
have some solution in this area. I hear a small number of people using lvm
snapshots and AFS.
But I don't really understand your comment about space inefficiency of
snapshots - in my mind, nothing could be more efficient, except to not have
snapshots (allow data deletion).
> > ...ZFS on linux. Apparently ZFS on linux has been working well now,
> > for at least a couple of years.
>
> We keep hearing rumors of that, but anyone actually using it?
I haven't personally used it, but I've heard it enough times that I've decided
I'm going to do it next time I need something like this. Literally the only
reason I use openindiana is to get a ZFS box, and I'd definitely prefer ubuntu
or centos.
> How about BtrFS now? I thought I saw some distributions switching to it
> as a primary FS.
It's probably ready. Around 3-ish years ago was the last time I tried it, and
it was *almost* ready then. Meaning, I built a server, and tested the
ever-loving hell out of it, and it passed all my tests. But then I put it into
production and we would occasionally see weird behaviors, and after a very time
consuming waste of effort spread over a few months, it was finally tracked down
to btrfs. So on that server we scrapped btrfs (and solved the problem), but it
was long enough ago that I wouldn't discourage trying again.
> I would *only* consider software RAID. So when I say RAID that's what I
> mean. I lump ZFS's RAID-Z with other software RAID, and don't consider
> it JBOD, as it is not using 100% of the storage for data.
Umm... I have a feeling you already know this, but the way you've phrased above
seems like maybe not? You definitely shouldn't lump zfs and btrfs in with
"other software raid," because the huge, major reason to use zfs/btrfs software
raid instead of hardware raid (besides system compatibility - ability to move
disks from one system to another) is the ability to detect & correct data
errors.
When the hardware presents only a single device to the OS, if a data error
occurs, then the OS has no way to tell the hardware "try reading the other
copy, to see if it's good." This means hardware JBOD and software raid are
necessary for the OS to do error correction. But many software raids (lvm, for
example) don't do checksumming and error correction.
> Now whether the overhead of RAID-Z is low enough that it makes more
> sense to use that over Ext4 on JBOD for a low-reliability backup pool is
> another matter.
This comment doesn't make any sense to me.
___
Discuss mailing list
Discuss@blu.org
http://lists.blu.org/mailman/listinfo/discuss