On Fri, Mar 6, 2015 at 9:38 AM, Gabriele Bartolini
<gabriele.bartolini@2ndquadrant.it> wrote:
> I believe the main point is to look at a user interface point of view.
> If/When we switch to a block level incremental support, this will be
> completely transparent to the end user, even if we start with a file-level
> approach with LSN check.
I don't think that's true. To have a real file-level incremental
backup you need the ability to take the incremental backup, and then
you also need the ability to take a full backup + an incremental
backup taken later and reassemble a full image of the cluster on which
you can run recovery. The means of doing that is going to be
different for an approach that only copies certain blocks vs. one that
copies whole files. Once we have the block-based approach, nobody
will ever use the file-based approach, so whatever code or tools we
write to do that will all be dead code, yet we'll still have to
support them for many years.
By the way, unless I'm missing something, this patch only seems to
include the code to construct an incremental backup, but no tools
whatsoever to do anything useful with it once you've got it. I think
that's 100% unacceptable. Users need to be able to manipulate
PostgreSQL backups using either standard operating system tools or
tools provided with PostgreSQL. Some people may prefer to use
something like repmgr or pitrtools or omniptr in addition, but that
shouldn't be a requirement for incremental backup to be usable.
Agile development is good, but that does not mean you can divide a big
project into arbitrarily small chunks. At some point the chunks are
too small to be sure that the overall direction is right, and/or
individually useless.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company