Re: chaining replication to setup a delta backup scheme

[Quanah Gibson-Mount <quanah@stanford.edu>]

--On Tuesday, November 11, 2003 7:04 PM -0500 james liang 
<blueyonder@excite.com> wrote:

> Couple of reasons.
>
> 1.  the daily change is very small but the total data size is huge.  to
> backup everything every day is wasteful.
>
> 2.  The primary has to be always on and has to handle a large volume of
> traffic.  I need to minimize any performance impact to production.
>
> 3.  I need multiple days of delta.  This way I can rollback everything up
> to the number of days of delta I keep.

We have quite a large DB.  However, by using slapcat to dump it to ldif and 
then the gzip, we end up with a fairly manageable size file (54M), and then 
we use an intelligent system to roll the dumps, so we have up to 5 days 
(but that is configurable) that we can drop back to.  Our daily change is 
rather large though.  I don't see what having your primary be always on has 
to do with whether or not you do a backup of the master, which I'd suggest 
keeping separate.

--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/TSS/Computing Systems
ITSS/TSS/Infrastructure Operations
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html