On 1/15/07, mike <mike503 at gmail.com> wrote:> 1) Is a hardware-based RAID behind the scenes needed? Can ZFS safely
> be considered a replacement for that? I assume that anything below the
> filesystem level in regards to redundancy could be an added bonus, but
> is it necessary at all?
ZFS is more reliable than a hardware RAID. OTOH, ZFS also relies on
the reliability of the underlying devices so ZFS benefits from
improvements to the former. Roch suggested using ZFS''s redundancy
features (i.e. mirror or raidz) regardless of underlying redundancy in
a conversation I had with him last year and I agreed with the
rational. ZFS''s checksumming and self-correction can save your data
even from hardware errors.
> 2) I am looking into building a 10-drive system using 750GB or 1TB
> SATA drives (when they come out) - I assume that I would need to plan
> for normal RAID-5 type storage usage with NUMDISKS-1 as the total
> amount of storage. Do I manage that at all, or is it completely
> managed by ZFS? If so, does ZFS calculate the free disk space after it
> takes in to account the space needed to store the parity data?
I''m not exactly sure what you are asking but ZFS will manage the disks
as whatever devices you specify them as. ZFS reports usable space
instead of raw.
> 3) I see RAID-Z is the equivalent of RAID-5, and RAID-Z2 is RAID-6
> (dual parity) - can these be interchanged (i.e. if a decision is made
> to switch to double the parity, can it be moved back to single parity?
> and vice-versa? assuming there is enough free space available for the
> second parity)
No, at least not now.
> 4) What happens if multiple disks fail? I am looking between RAID-Z
> and RAID-Z2. I understand one or two could die in Z or Z2, what
> happens if half the array disappears (one of the enclosures power
> supplies fails, perhaps) - is there some method to "halt" until
it is
> manually fixed then?
The system panics for now. There was a discussion about this if you
search the list archives.
> 5) Follow up question: can a new, larger disk be put in it''s
place?
> What about smaller? Will ZFS understand that?
You cannot replace a disk with a smaller disk.
> 6) Assumption: disks can be permanently removed (as long as there is
> enough leftover space)
Not now. Not sure if this will change.
> 7) Will upgrades to the filesystem work easily? When encryption is
> supported I would be interested in upgrading to that version to take
> advantage of that.
zpool upgrade.
> Sorry for the amount of questions, it will help me gain more
> understanding of this. I''m trying to consume as much information
about
> ZFS as possible, and it helps to get the official answers back with
> questions phrased in a format I understand.
>
> Thanks in advance!
>
> P.S. Just for information''s sake, I plan on setting up one or
multiple
> sets of 5 or 10 drive eSATA-based arrays - for instance the 10 bay, 2
> port eSATA one here:
> http://fwdepot.com/thestore/product_info.php/products_id/1578
> alongside some Solaris x86/OpenSolaris-compliant eSATA adapter [if
> anyone has any suggestions, feel free to reply directly to me!]) -
> this would be used for home media storage + offsite redundant backup
> for multiple locations for my webhosting business. Not heavy traffic,
> perhaps 3-5 people accessing it at once. Files probably on average
> more than a few megs - so wouldn''t be *too* sparse...
> _______________________________________________
> zfs-discuss mailing list
> zfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
>
--
Just me,
Wire ...