Migration from the KQ Implementation

Gordan Bobic gordan.bobic at gmail.com
Thu May 5 16:34:03 EDT 2011

On 05/05/2011 02:56 PM, Steve Costaras wrote:
> While I don't don a beard (grey or otherwise), and completely understand
> the need for recovery or alternative means to pull data from a FS to
> handle unexpected situations I think it should be clear that ZFS is
> actualy not much worse than the aggregate of it's comprising functions.
> Comparing it to a traditional file system (ufs, jfs, xfs, ext[1-4], et
> al) is not really apples to apples.
> ZFS in essence is three items in one, a RAID layer (regardless if this
> is different than block type traditional raid it is doing the same
> function) so that can be compared to say metadevices (MD; svm; etc.); a
> logical volume manager (VxVM; lvm; svm; etc. ) and then your FS of
> choice. You are getting all three layers basically merged together so
> failures in any one of the functional aspects can all have the same
> effect (data not available). Add to this other options that make it even
> harder such as dedup and compression. Traditionally, each 'layer' could
> be operated on one at a time so you can address lower issues before
> getting to the higher functions.

I don't agree with this. Although the features are merged for reasons of 
performance, logically they are still separate, and repairing a pool (MD 
equivalent) should be doable independently of higher level functions.

> Remember what ZFS is (data/block availability and integrity) but not a
> solution for backups.

It's not a replacement for backups, but backups should only be required 
in case of either extreme hardware failure or user error (deleting 
files). The latter can be side-steppable by using something like CopyFS.

fsck is required for the cases inbetween hardware failure and user error.

One thing we have to remember is that advanced file systems like ZFS are 
usually used in cases where the amount of data is very large. That makes 
availability of good recovery tools more important because restoring 
from backups will take a prohibitively long time compared to fsck. Most 
people (myself included) do not fully grok just how much data 10TB is 
until you actually have to do something with _all_ of it (e.g. copy it 
from a backup server).


More information about the zfs-discuss mailing list