Expand Raid-Z (expand_o_matic_raid_z)

[mrsieb]

So only wondering if there are news in this case ? Somebody know more about that?

-> http://blogs.oracle.com/ahl/entry/expand_o_matic_raid_z

 

[ProtoSD]

Nice article. This kind of change is not likely to come from the FreeNAS developers, but will have to wait until it gets implemented in FreeBSD first. Since it's a feature in such demand and a basic path has been described, it's probably more likely it will start being worked on by someone sooner than later. Thanks for sharing that link.

 

[louisk]

I'm probably in the minority here, but I'm of the opinion that growing by adding a vdev, and the ability to replace disks with larger disks to grow an existing vdev is sufficient.

 

[mrsieb]

I know that this kind of „feature“ must be implemented in FreeBSD not by FreeNas at first in zfs by sun ;-)
Any how this is one of the missing link for home use maybe special for multimedia content.

“ It's common for a home user to want to increase his total storage capacity by a disk or two at a time,
but enterprise customers typically want to grow by multiple terabytes at once so adding on a new RAID-Z stripe isn't an issue.”

Let me explain it this way:

As Home User you by a couple of Disk, if you can afford for example 4x2TB =8TB
So now you think about how to handle this 8TBs to make a good balance between
disk fail and disk space (speed is not really interesting in this case)
At least you choose some kind of Raid-5 now you still have a backup problem.
Because to backup 8TB Raid 5 at least you need additionally 6TB space how every on which media.
So I guess for home user this is not possible under normal conditions. So you try to be protected against disk fail.
Or in other words. Keep data lose small as possible in worst case. The point is on every raid system the data is
striped over the mirror. Two things happen. First it is not possible to restore any data from a single disk and you
cannot expand the mirror with a single disk without rebuild the raid and loose all data expected you can backup it
in the mean time So if you want to have more space you have to by 4 bigger disk or additionally 4 disk with same
space again you create raid5 now you have 12TB storage but split it up in 2x4Mirros each one parity. You got space
like raid-5 parity 2 (or in case zfs raidz2) but not the safety of the disk fail because instead it would be better to build
a raid5-parity2 over all 8 disk or raidz-2 or parity1 or z1 to get 2TB more space….and again this is not possible...
excepted you can backup your data in the mean time..

uhh.. hope somebody can follow

greetings

 

[bubulein]

Moin,

may I suggest some kind of disklayout which is ok for most homeuser.
I call it "Heaven & Hell"
The basic idea behind it is to wok with two arrays.
Heaven is a mirror and Hell is a stripe.From the NAS point of few all data on Heaven is more or less save, while loosing a disk on the other array take you to "hell".

In my case Hell contain all my media-files. For each TB I have an external disk which backup this files. All new files are put onto "Heaven", and once a week or month I cleanup Heaven.
So..
Copy files from Heaven to Hell
Backup Hell
Test Backup
Delete Heaven.
Since Heaven has 1TB here I usually have much time to find new drives (or call it save some money).
Compared to a raidz I can easy upgrade my pool by adding an additional drive. Since Heaven only need to be a small mirror I can use small fast drives to improve write speed on my NAS. If you like you may setup a rsyncjob for Heaven, and mirror the files to Hell hourly etc. so you can save an additional drive while still having a very basic Fail Over.

-Christian