I build cloud-based systems for startups and enterprises. My background in operations gives me a unique focus on writing observable, reliable software and automating maintenance work.
I love learning and teaching about Amazon Web Services, automation tools such as Ansible, and the serverless ecosystem. I most often write code in Python, TypeScript, and Rust.
B.S. Applied Networking and Systems Administration, minor in Software Engineering from Rochester Institute of Technology.
There are certain topics in BIND9 that are well documented, like running a slave DNS server, or delegating a subzone. One of the least documented setups I’ve seen also happens to be one I had to implement.
I found myself in need of redundant split-horizon name servers behind a static NAT. I went through several iterations building this topology, and I learned something new at every step.
The topology would have three zones:
The tricky bit was something I didn’t know about some versions of Cisco IOS: that it tries to “help” people who are running DNS servers inside a NAT. It goes ahead and rewrites DNS responses. This is nice if you’re setting up a simple topology with forward lookups only. When you begin doing reverse lookups, the automatic rewriting can become a non-trivial problem. It is inconsistent in which query types it will rewrite.
For example, a request for a PTR record from a reverse zone such as PTR 72.0.10.9
will get rewritten as PTR 10.100.100.9
by the router. This causes
a problem when queries against your external address can resolve as
internal-only names, reverse records can leak outside via both the primary and
secondary name servers.
I posted the named.conf and associated zone files on Github for reference. On their own they don’t make for good reading, but I’ll be referring to them throughout this post.
The concept of split-horizon is well documented, and the internal-view
and
external-view
in etc/named.conf are relatively straightforward.
Generally, you’d only do this if you had a 1:1 NAT for some of your devices. In
my case, the mail server (among other machines) had to be available inside and
outside the NAT.