[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: fs block level syncing



Paul Raines <raines nmr mgh harvard edu> writes:

> On Fri, 13 Feb 2004, Sean Neakums wrote:
>> Off the top of my head I was thinking that maybe you could do
>> something with lvm snapshots and dump, doing a restore on the remote
>> side, but I have no idea how practical and robust that would be.
>
> I looked into that but LVM/EVMS snapshots store only the difference.  What
> I need is complete other copy so if the original paritions dies completely,
> I still have everything (at least up to the last mirror job).

Right, but the idea of dumping the snapshot is that you will be
dumping a consistent image of the filesystem, which can then be
restores on the other side and used safely.  (At least, I think that
snapshots give you a consistent view of the FS.)  Once the dump is
complete, you ditch the snapshot.

> But you got me thinking about dump/restore.  Does dump/restore do things at
> the block level?  If I have two equal partitions and I make my initial
> image doing a dump on the first and piping it into a restore one the
> second. Then at later intervals do an incremental dump on the first piping
> it again into a restore on the second, will it just do the most 
> efficient thing (i.e. handle renames as only a change to the directories
> involved)?  
>
> I am not too familiar with dump so don't know if incremental dumps are 
> possible in such a scheme.

I'm very rusty on dump, but if I recall correctly, one of the
arguments is the "level", which will dump only the differences since
the last dump at level N - 1.  Level 0 is special and denotes a full
dump.




[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]