zfs discuss - Mar 2009 - is ''zfs receive'' atomic per snapshot?

Can we assume that any snapshot listed by either ''zfs list -t
snapshot'' or
''ls .zfs/snapshot'' and previously created with ''zfs
receive'' is complete and
correct? Or is it possible for a ''zfs receive'' command to fail
(corrupt/truncated stream, sigpipe, etc...) and a corrupt or incomplete
snapshot to exist after the command failed? Thoughts on this would be much
appreciated.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090319/dc2334e4/attachment.html>

Jos? Gomes wrote:
> Can we assume that any snapshot listed by either ''zfs list -t 
> snapshot'' or ''ls .zfs/snapshot'' and previously
created with ''zfs
> receive'' is complete and correct? Or is it possible for a
''zfs
> receive'' command to fail (corrupt/truncated stream, sigpipe,
etc...)
> and a corrupt or incomplete snapshot to exist after the command 
> failed? Thoughts on this would be much appreciated.
receives are all-or-nothing per snapshot.
 -- richard

Jos? Gomes wrote:
> Can we assume that any snapshot listed by either ''zfs list -t
snapshot''
> or ''ls .zfs/snapshot'' and previously created with
''zfs receive'' is
> complete and correct? Or is it possible for a ''zfs
receive'' command to
> fail (corrupt/truncated stream, sigpipe, etc...) and a corrupt or 
> incomplete snapshot to exist after the command failed? Thoughts on this 
> would be much appreciated.
If a ''zfs recv'' fails for any reason (bad stream, system
crash, etc), its
space will be reclaimed.  The snapshot won''t appear in the namespace
until it
has been successfully received.

--matt

2009/3/19 Richard Elling <richard.elling at gmail.com>
> Jos? Gomes wrote:
>
>> Can we assume that any snapshot listed by either ''zfs list -t
snapshot'' or
>> ''ls .zfs/snapshot'' and previously created with
''zfs receive'' is complete and
>> correct? Or is it possible for a ''zfs receive''
command to fail
>> (corrupt/truncated stream, sigpipe, etc...) and a corrupt or incomplete
>> snapshot to exist after the command failed? Thoughts on this would be
much
>> appreciated.
>>
>
> receives are all-or-nothing per snapshot.
> -- richard
>
>
This is good. Thanks a lot, Richard.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090320/b04445ad/attachment.html>

2009/3/19 Matthew Ahrens <Matthew.Ahrens at sun.com>
> Jos? Gomes wrote:
>
>> Can we assume that any snapshot listed by either ''zfs list -t
snapshot'' or
>> ''ls .zfs/snapshot'' and previously created with
''zfs receive'' is complete and
>> correct? Or is it possible for a ''zfs receive''
command to fail
>> (corrupt/truncated stream, sigpipe, etc...) and a corrupt or incomplete
>> snapshot to exist after the command failed? Thoughts on this would be
much
>> appreciated.
>>
>
> If a ''zfs recv'' fails for any reason (bad stream, system
crash, etc), its
> space will be reclaimed.  The snapshot won''t appear in the
namespace until
> it has been successfully received.
>
> --matt
>
Great. Thank you, Matthew.
J.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20090320/46b52027/attachment.html>