[prev in list] [next in list] [prev in thread] [next in thread] 

List:       djbdns
Subject:    Re: Full featured djbdns jumbo patch
From:       Andy Dustman <andy () dustman ! net>
Date:       2001-10-11 18:29:52
[Download RAW message or body]

On Thu, 2001-10-11 at 04:38, D. J. Bernstein wrote:
> >     - add support to tinydns-data to load multiple data files
> >     specified on command line
> 
> Merging multiple AXFR results properly is more complicated than this. It
> should be handled by a separate tool. Note that sites with many zones
> can't even fit the list into the command line.

A better solution would be: If data is a directory, traverse the
directory tree, reading each file, and using it's timestamp for the
default SOA serial number for all zones in that file. Then AXFR tools
can update individual files within the data directory. To determine if
data.cdb needs to be rebuilt, check the mtime on all files within data
and it's subdirectories. Once one is found to be newer than data.cdb,
rebuild; otherwise leave it alone.

This will take longer than reading a simple data file, but then,
building data from a bunch of separate files takes time, too. May be a
worthwhile trade-off in some cases. Case in point: You have ten thousand
zones you are secondary for. To make a single data file, you have to cat
ten thousand files together, write it to disk, then run tinydns-data to
make data.cdb. If data is very large, tinydns-data will probably not be
able to get much of it out of caches, so all the data is read twice, and
written twice (once as data, once as data.cdb). OTOH, if tinydns-data
can read the individual files directly, they only have to be read and
written once.

For many setups, of course, a single data file works fine.

-- 
Andy Dustman         PGP: 0x930B8AB6
    @       .net     http://dustman.net/andy
You can have my keys when you pry them from my dead, cold neurons.

[prev in list] [next in list] [prev in thread] [next in thread]