I was asking exactly for what you've written. Thanks god one actually
understood my original question, this is due to my poor english. I would
current on-disk structures.
Post by Sam Zaydel via illumos-zfsIf I were to suggest anything it might be that a convenient feature
would be an implicit restore by creating a link for example to a
file/dir in a snapshot. Basically, if you were to create a link in
some place to a file in a snapshot you could get a read/write clone of
that file/directory. This is by no means a trivial thing to do, but I
think this is sort of what you are asking for, maybe?
On Mon, Aug 18, 2014 at 8:16 AM, Richard Kojedzinszky via illumos-zfs
That is why I asked here, and not on an -fs list. Basically, some
Post by Richard Kojedzinszky via illumos-zfsZFS expert could think about it if it is possible at ZFS
layer to make that link, to make a directory entry pointing to an
object which exists in a snapshot. Of course, this wont be a
POSIX fileop, as a regular hardlink cant be mode cross-device. That
is why I asked here the ZFS guys, developers what do they
think about the problem. For me, with that I dont know much of ZFS,
I think the data structures may allow such an operation.
My question is that would it be hard to implement such
functionality or not?
Regards,
Ignoring any other logistical problems which may exist...
If it were a hardlink, it'd be a read-only file.
What would it mean to have a single read-only file on an otherwise
read-write filesystem?
If you can modify it and the original doesn't change, it's not really
a hardlink.
If you can't modify it, it's not really useful in most cases
compared
to just symlinking to the RO copy...
- Rich
On Mon, Aug 18, 2014 at 8:08 AM, Richard Kojedzinszky via
illumos-zfs
Actually I intentionally wrote just 'ln', and not 'ln -s'. Somehow a
hard-link could be created for that existing file in the snapshot. Would it
need a big change in zfs code, big redesign, or is it possible in theory to
implement it?
Thanks
18 авгÑÑÑа 2014 г. 9:03:04 CEST, Richard Kojedzinszky via
illumos-zfs
Dear all,
Thanks for the comments. Actually, we have lz4 enabled on our dataset,
so zfs will make the holes holes again when copying the data back.
The original question is more about theory, the original file was also
on the dataset which the snapshot was taken of, so basically a
# ln .zfs/snapshot/snap/file ./file
like operation could be accomplished somehow, or is it more complex to
achieve? I dont know zfs internals, but this could be a new entry
in
the
dir pointing to the same file (object) in the dataset which existed
before.
Regards,
This is why I use qcow2 :(
Okay, so setting any compression should compress 0 blocks to take no
actual space.
If the files are sparse, Linux cp has efficient handling of sparse files.
If VMware isn't using either files with a dynamic size by default, or
sparse files at a bare minimum, please for a bug report with
upstream.
On Aug 17, 2014 2:46 PM, "Sam Zaydel via illumos-zfs"
If vmware, chances are, re-thinning will be necessary, since
recovering VM from snapshot will cause it to take up entire amount
of disk originally allocated, assuming it was thin provisioned of
course.
On Sun, Aug 17, 2014 at 12:03 PM, Richard Elling via illumos-zfs
On Aug 17, 2014, at 11:24 AM, Richard Kojedzinszky
I've forget to mention that weh have all images on a single
dataset, so clone/promote to an older snapshot is not reasonable.
I'm sure by now you realize the folly of this decision :-) But
perhaps you are running vmware where such configs
are more common.
We would need to revert only one file of the snapshot.
Any ideas?
clone still works, just point the one VM at the clone.
-- richard
Kojedzinszky Richard
On Aug 17, 2014, at 2:05 AM, "Richard Kojedzinszky via
Dear ZFS devs,
Post by Richard Kojedzinszky via illumos-zfsWe have a simple VPS hosting setup, where a ZFS dataset is
exported to NFS, hosting the VM's images. This dataset is being
snapshotted regularly. Demand has come to recover only one VM's
image: I mounted the snapshot, then copied the file from snapshot
over its original place. I dont know, but in theory would it be
possible with a simple zfs command to just make a link somehow to
the file in the snapshot at the original place? That way the
recovery would be very fast, there would be no need to read the
whole file (which can be full of holes), etc.
Could this feature be implemented in ZFS?
Post by Richard Kojedzinszky via illumos-zfssee clone
-- richard
Regards,
Kojedzinszky Richard
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
[1]
https://www.listbox.com/member/archive/rss/182191/22820713-4fad4b89
[6]
[2]
Post by Richard Kojedzinszky via illumos-zfsModify Your Subscription: https://www.listbox.com/member/?& [2]
[3]
Powered by Listbox: http://www.listbox.com [3] [4]
Post by Richard Kojedzinszky via illumos-zfs-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
[1]
https://www.listbox.com/member/archive/rss/182191/25402478-0858cafa
[7]
[5]
Post by Richard Kojedzinszky via illumos-zfsModify Your Subscription: https://www.listbox.com/member/?& [2]
[3]
Powered by Listbox: http://www.listbox.com [3] [4]
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1] [1]
https://www.listbox.com/member/archive/rss/182191/24342081-7731472e
[8]
[6]
Post by Richard Kojedzinszky via illumos-zfsModify Your Subscription: https://www.listbox.com/member/?& [2] [3]
Powered by Listbox: http://www.listbox.com [3] [4]
--
Join the geek side, we have Ï!
Please feel free to connect with me on LinkedIn.
http://www.linkedin.com/in/samzaydel [5] [7]
ILLUMOS-ZFS | Archives [1] [8] | Modify [9] Your Subscription [4]
------
[1] https://www.listbox.com/member/archive/182191/=now [1]
[2]
https://www.listbox.com/member/archive/rss/182191/22820713-4fad4b89
[6]
[3] https://www.listbox.com/member/?& [4];
https://www.listbox.com/member/archive/rss/182191/25402478-0858cafa
[7]
[6]
https://www.listbox.com/member/archive/rss/182191/24342081-7731472e
[8]
[7] http://www.linkedin.com/in/samzaydel [5]
https://www.listbox.com/member/archive/rss/182191/23054485-60ad043a
[9]
[9]
While the symlink to the snapshot virtual directory might be possible
(usability subject to dataset access rights delegation, see zfs
allow), this would be a read-only file. You might copy it back into
the 'live' dataset (or rather rsync --inplace so as to not waste space
during the 'recovery'), you won't use it well for a live VM. If
dealing with such symlinks, also use zfs hold to keep the snapshot
from disappearing.
Hth,
//Jim
--
Typos courtesy of K-9 Mail on my Samsung Android
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
https://www.listbox.com/member/archive/rss/182191/25402478-0858cafa
[7]
https://www.listbox.com/member/?& [2]
Powered by Listbox: http://www.listbox.com [3]
--
Richard Kojedzinszky
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
https://www.listbox.com/member/archive/rss/182191/24536931-5d25148d
[10]
https://www.listbox.com/member/?& [2]
Powered by Listbox: http://www.listbox.com [3]
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
https://www.listbox.com/member/archive/rss/182191/25402478-0858cafa
[7]
https://www.listbox.com/member/?& [2]
Powered by Listbox: http://www.listbox.com [3]
--
Richard Kojedzinszky
-------------------------------------------
illumos-zfs
Archives: https://www.listbox.com/member/archive/182191/=now [1]
https://www.listbox.com/member/archive/rss/182191/24342081-7731472e
[8]
https://www.listbox.com/member/?& [2]
Powered by Listbox: http://www.listbox.com [3]
--
Join the geek side, we have Ï!
Please feel free to connect with me on LinkedIn.
http://www.linkedin.com/in/samzaydel [5]
ILLUMOS-ZFS | Archives [1] [7] | Modify [11] Your Subscription [3]
------
[1] https://www.listbox.com/member/archive/182191/=now
[2] https://www.listbox.com/member/?&
[3] http://www.listbox.com
[4] https://www.listbox.com/member/?&amp
<https://www.listbox.com/member/?&>
[5] http://www.linkedin.com/in/samzaydel
[6] https://www.listbox.com/member/archive/rss/182191/22820713-4fad4b89
[7] https://www.listbox.com/member/archive/rss/182191/25402478-0858cafa
[8] https://www.listbox.com/member/archive/rss/182191/24342081-7731472e
[9] https://www.listbox.com/member/archive/rss/182191/23054485-60ad043a
[10] https://www.listbox.com/member/archive/rss/182191/24536931-5d25148d
[11]
https://www.listbox.com/member/?&id_
secret=25402478-a0052abc