|Main Archive Page > Month Archives > bind-users archives|
> We use bind for DNS.
> At the moment we have one primary server that delegates updates to it`s two slave servers.
> Since everything nowadays is dependant on DNS I would like to cluster my primary server in case of a hardware failure or error.
> So, how do I setup two primary bind servers that keep each other in sync one way or the other.
> I`ve been surfing the internet, but couldn`t find any satisfactory solution.
I've never tried this, but I think it might work:
The hard part of this is the dynamic DNS, so looking at that part of it
Use multiple slave servers in an anycast architecture for the published
MNAME server, and use update forwarding on the slave servers to replicate
DDNS to multiple hidden masters. The slaves would then all zone-transfer
the info back from the hidden masters.
To conserve hardware, you can have 2 instances of BIND on a single server
so that each slave has a hidden master on the same box.
The key to anycasting is that if a slave server stops working, the host
route to that server must be removed from the router asap. This can be
done best using Cisco routers with static routes that track "ip sla"
monitors configured to check DNS. Otherwise it can be done using OSPF
between the server and the router with self-monitoring scripts on the
servers that take down the loopback interface if the service is down.
For the non-dynamic DNS, you can have all masters share a common NAS
directory for non-dynamic zone files as well as the shared conf file
-- but keep the dynamic (and slave) data in local directories. Any time
the non-dynamic data changes, you will need to "kill -1" all of the BIND
processes configured as master for the affected zones. This can be
automated with scripts using rndc.
I'd be glad to help further, but more importantly I'd like to know how
well it works for you before I dare try it myself.
-- Gordon A. Lang / 313-819-7978
bind-users mailing list