Amanda-Users

Re: LVM snapshots

2006-07-07 17:00:24
Subject: Re: LVM snapshots
From: Jon LaBadie <jon AT jgcomp DOT com>
To: amanda-users AT amanda DOT org
Date: Fri, 7 Jul 2006 16:52:04 -0400
On Fri, Jul 07, 2006 at 12:47:15PM -0600, Michael Loftis wrote:
> 
> 
> >Does anyone use or have knowledge of using LVM snapshots with Amanda
> >backups?
> >
...
> >I'm just wondering what happens during the freeze - how freezing "all
> >activity to and from the filesystem to reduce the risk of problems"
> >affects the system?  One would imagine that disk writes are somehow
> >queued up and complete when the file system is unfreezed again?
> 
> It's only stopped long enough to make the bitmap table.  After that it's 
> Copy On Write.  Meaning a snapshot is free until the 'original' starts to 
> differ then the snapshot starts to take up space because you have to make 
> copies of the original blocks for the snapshot once they start to change.
> 

Apologies for an OT comments;  This seems like an ineffecient way to
implement snapshots.  Granted the inode for a changed file has to be
copied, but I don't see why the data blocks do.  If I'm changing a
data block, I have to write that block to disk anyway.  Whether I
overwrite the original block, or a newly allocated one doesn't matter.
If the original was just left in place, the snapshot could continue
to use it.  Its copied inode would already have the pointers to the
original blocks.  Otherwise you have to copy the inode, allocate a
new block for the copy of the original block, change the copied inode
to point to the copied block, and still have to write the changed data
for the active file.


-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)

<Prev in Thread] Current Thread [Next in Thread>