Failsafe Replication

[Max Clark-2]

Hello,

I have been evaluating alternate DNS topologies to our current
configuration. Ideally I would prefer an architecture where the master
is hidden, and that the public slaves can run independent without
dependance on an external process (i.e. mysql).

Initially I was under the impression that the supermaster would
accomplish this for me, but based on the docs I've read it appears
that the superslave requires a sql backend to function properly.

Reverting back to the bind backend requires an update of the
configuration file and periodic reloads of the process to see and act
on changes to the zones. Based on experience with BIND this is
sub-optimal over the long run.

MySQL replication provides "native" configuration and fast replication
for pdns - however it adds an additional layer of complexity,
software, process to the mix that adds another failure point and
requires additional monitoring and management.

Is there another technique / configuration that I am missing here?

Thanks,
Max
_______________________________________________
Pdns-users mailing list
Pdns-users@...
http://mailman.powerdns.com/mailman/listinfo/pdns-users

[Ask Bjørn Hansen]

On Aug 24, 2009, at 19:44, Max Clark wrote:

> I have been evaluating alternate DNS topologies to our current
> configuration. Ideally I would prefer an architecture where the master
> is hidden, and that the public slaves can run independent without
> dependance on an external process (i.e. mysql).
>
> Initially I was under the impression that the supermaster would
> accomplish this for me, but based on the docs I've read it appears
> that the superslave requires a sql backend to function properly.

Hi Max,

The generic sqlite backend supports superslave operation and won't  
require an external process running.

That being said - using MySQL replication really is very nice  
operationally if you have some experience running mysql with  
replication.  For example:

  - It makes it easy to clean up old data
  - There are good tools for verifying the consistency of the data  
(maatkit's mk-table-checksum)
  - You can have replication happen over SSL
  - For large zones you'll only be transferring the changes rather  
than the full zone
  - There are lots of tools to monitor that MySQL replication is  
working properly
  - It's easy and fast to reload a failed replication slave with the  
master data

You are right that, of course, it does mean you need to keep one more  
process running; but to operate even a single purpose server you  
usually also need some set of sshd, ntpd, crond, some MTA, syslogd,  
some daemon to collect system statistics, etc etc.


  - ask

--
http://develooper.com/ - http://askask.com/


_______________________________________________
Pdns-users mailing list
Pdns-users@...
http://mailman.powerdns.com/mailman/listinfo/pdns-users