Discussion:
Snapshot rebase?
Schlacta, Christ
2013-11-21 22:21:23 UTC
Permalink
Is it possible to rebase a snapshot from one ancestor to another copy of
itself? Or one filesystem on another ancestor snapshot of itself?
Example:

zfs snapshot tank/***@1;
zfs send tank/***@1 | zfs recv tank/shark;
zfs rebase tank/fish tank/shark

If this is possible, what does the exact process look like? If it's not
possible, is it feasible to implement?
If it's implementable, can it be done online, as in with no downtime or
unmount needed?

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org
Durval Menezes
2013-11-21 22:33:26 UTC
Permalink
Hi Christ,
Post by Schlacta, Christ
Is it possible to rebase a snapshot from one ancestor to another copy of
itself? Or one filesystem on another ancestor snapshot of itself?
Post by Schlacta, Christ
zfs rebase tank/fish tank/shark
Sorry if I'm being too dense, but exactly what would the last command above
do to the two pools and/or the snapshot of the first one?

Cheers,
--
Durval.

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org
Schlacta, Christ
2013-11-21 22:53:50 UTC
Permalink
Post by Durval Menezes
Hi Christ,
Post by Schlacta, Christ
Is it possible to rebase a snapshot from one ancestor to another copy
of itself? Or one filesystem on another ancestor snapshot of itself?
Post by Durval Menezes
Post by Schlacta, Christ
zfs rebase tank/fish tank/shark
Sorry if I'm being too dense, but exactly what would the last command
above do to the two pools and/or the snapshot of the first one?
The last command does exactly what rebase does with virtual machine disk
images. The data in tank/fish is now represented relative to tank/shark
instead of relative to tank/fish. The exact data contained therein remains
logically consistent. In theory, there should be no down time required if
the mount point is set manually allowing file operations to continue across
the rebase operation.

It obviously is only well defined for file systems on the same pool.
Post by Durval Menezes
Cheers,
--
Durval.
To unsubscribe from this group and stop receiving emails from it, send an
email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org
Massimo Maggi
2013-11-21 23:21:13 UTC
Permalink
Are you looking for "zfs clone / zfs promote" ?
Post by Schlacta, Christ
Post by Durval Menezes
Hi Christ,
Post by Schlacta, Christ
Is it possible to rebase a snapshot from one ancestor to another copy of
itself? Or one filesystem on another ancestor snapshot of itself?
zfs rebase tank/fish tank/shark
Sorry if I'm being too dense, but exactly what would the last command
above do to the two pools and/or the snapshot of the first one?
The last command does exactly what rebase does with virtual machine disk
images. The data in tank/fish is now represented relative to tank/shark
instead of relative to tank/fish. The exact data contained therein remains
logically consistent. In theory, there should be no down time required if
the mount point is set manually allowing file operations to continue across
the rebase operation.
It obviously is only well defined for file systems on the same pool.
Post by Durval Menezes
Cheers,
--
Durval.
To unsubscribe from this group and stop receiving emails from it, send an
To unsubscribe from this group and stop receiving emails from it, send an
To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org
Schlacta, Christ
2013-11-21 23:42:59 UTC
Permalink
Maybe, but not quite. My use case is to use zfs send to effect reordering
of data and application of modified permissions (hash, compress) on live
data sets. Some data sets it's easy to take offline, but I'd prefer to not
have to take the entire vm store down.
Post by Massimo Maggi
Are you looking for "zfs clone / zfs promote" ?
Post by Schlacta, Christ
Post by Durval Menezes
Hi Christ,
Post by Schlacta, Christ
Is it possible to rebase a snapshot from one ancestor to another copy
of
Post by Schlacta, Christ
Post by Durval Menezes
Post by Schlacta, Christ
itself? Or one filesystem on another ancestor snapshot of itself?
zfs rebase tank/fish tank/shark
Sorry if I'm being too dense, but exactly what would the last command
above do to the two pools and/or the snapshot of the first one?
The last command does exactly what rebase does with virtual machine disk
images. The data in tank/fish is now represented relative to tank/shark
instead of relative to tank/fish. The exact data contained therein
remains
Post by Schlacta, Christ
logically consistent. In theory, there should be no down time required if
the mount point is set manually allowing file operations to continue
across
Post by Schlacta, Christ
the rebase operation.
It obviously is only well defined for file systems on the same pool.
Post by Durval Menezes
Cheers,
--
Durval.
To unsubscribe from this group and stop receiving emails from it, send
an
Post by Schlacta, Christ
To unsubscribe from this group and stop receiving emails from it, send an
To unsubscribe from this group and stop receiving emails from it, send an
To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe-VKpPRiiRko7s4Z89Ie/***@public.gmane.org
Loading...