Hi Tomasz,
On 2010-07-09, at 10:26 AM, Tomasz Chmielewski wrote:
> Hi,
>
> I'm about to set up bind with GeoIP patches.
>
> What I'm not sure, is how do you guys handle high availability?
>
> Suppose I have zones for Americas and Europe, and a destination server in
> Europe dies - how do you handle it so that new (i.e. web) requests hit
> American servers only?
>
> Set TTL to low values (i.e. 10 minutes max) and reconfigure the zones if
> European servers are down? Something else?
>
> I assume typical hosting, without access to sophisticated network settings,
> like BGP.
I think that you're confusing two issues. DNS has redundancy baked in.
Let's say you have 2 nameservers
ns-europe.example.com ( which is physically located in North America )
ns-americas.example.com ( which is physically located in Europe )
and both of those are authoritative for this zone
geoip.example.com
Now that's a special zone being operated using the GeoIP patches so that if
either of the nameserver receives a query from a European IP it gets a custom
answer suitable for Europeans and if it gets a query from an African IP it
returns a custom answer suitable for Africans.
Both of your nameserves answer the same way. European queries do not go
exclusively to the European server. If ns-europe is down resolvers will get an
answer from ns-america instead.
This situation does involve a failed query to the down server, followed by a
retry and is probably Good Enough. If you absolutely must avoid the penalty of
a timeout + retry you need to make the individual nameservers redundant and
that probably means anycast.
dave
_______________________________________________
bind-users mailing list
bind-users@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users
