Redundancy - load balancing
kcd at daimlerchrysler.com
Mon Oct 21 23:38:40 UTC 2002
> A few quick questions about redundancy..
> Sceario: A company has 2 ISP's, with a different ip-range for each ISP. Both
> ranges is terminated in separate routers, which NAT's the official addresses
> to private ones on the inside. One name server is connected to each link.
> Preferably, One would like resolvers on the internet to recieve answers to
> all queries from ns1 (and thus, use the addresses/link from ISP1 for data
> transfer), but if the link from ISP1 fails, resolvers should query ns2, and
> thus, use the adresses/link from ISP2, and thereby provide redundancy for
> incoming traffic.
> setup in db.test.com:
> test.com. IN SOA ns1.test.com. blabla.test.com ( <removed options> )
> test.com. IN NS ns1.test.com
> test.com. IN NS ns2.test.com
> ns1.test.com IN A 10.1.1.1
> ns2.test.com IN A 192.168.1.1
> Would this work?
> Would resolvers always query ns1, and if they are not able to connect to
> ns1, query ns2, or would this setup provide round-robin load-balancing
> instead (or not work at all..)?
> any comments on this would be greatly appreciated.
You don't need to do anything special for load-balancing and redundancy of
nameservers. Since both of those servers are in the NS records for the zone,
other nameservers will spread out the query load by keeping track of the
RTT (round-trip time) of queries and responses and preferring the
"faster" nameserver. If one of the nameservers should fail, then the RTT will
quickly get so penalized that other nameservers will stop using it for a while.
More information about the bind-users