What happens when... web server with 2 internet connections

Started by Dieselboy, August 14, 2019, 08:00:49 AM

Previous topic - Next topic

Dieselboy

I have a web application server for internal users, accessible publicly via dns.name.com. This resolves to the NAT IP via the main / primary internet link for the site. The site has a 2nd internet connection and the server has a 2nd NAT and can be publicly reached via the 2nd internet connection.

If I resolve dns.name.com to both IP addresses, is the inbound traffic round-robin? What other issues are there with this other than if one connection should fail, incoming requests to that IP will take a long time without response before the browser tries the other IP? Or am I mistaken at this point? I think the caveat is both connections need to be up for a responsive service, ie it's no good having a cold standby IP being resolved because occasionally it would be tried resulting in delays and retries.

deanwebb

So, once a device gets its DNS request satisfied, it'll cache that and use it for the whole connection. Round-robin DNS will spread the connections across the IP addresses, but not necessarily the load.
Take a baseball bat and trash all the routers, shout out "IT'S A NETWORK PROBLEM NOW, SUCKERS!" and then peel out of the parking lot in your Ferrari.
"The world could perish if people only worked on things that were easy to handle." -- Vladimir Savchenko
Вопросы есть? Вопросов нет! | BCEB: Belkin Certified Expert Baffler | "Plan B is Plan A with an element of panic." -- John Clarke
Accounting is architecture, remember that!
Air gaps are high-latency Internet connections.

Dieselboy

Thanks for the clarification. So I should be fine to publish the web app in this way and allow connections to chose whichever inbound the browser wishes to use at that time. I dont mind having 80% on one connection or the other for example.

deanwebb

Quote from: Dieselboy on August 14, 2019, 09:36:24 PM
Thanks for the clarification. So I should be fine to publish the web app in this way and allow connections to chose whichever inbound the browser wishes to use at that time. I dont mind having 80% on one connection or the other for example.

Pretty much, yeah. If you actually needed load balancing over dozens of addresses because of ecommerce stuff, then there's a more sophisticated solution you could go and buy.
Take a baseball bat and trash all the routers, shout out "IT'S A NETWORK PROBLEM NOW, SUCKERS!" and then peel out of the parking lot in your Ferrari.
"The world could perish if people only worked on things that were easy to handle." -- Vladimir Savchenko
Вопросы есть? Вопросов нет! | BCEB: Belkin Certified Expert Baffler | "Plan B is Plan A with an element of panic." -- John Clarke
Accounting is architecture, remember that!
Air gaps are high-latency Internet connections.

Otanx

Also pay attention to DNS TTL. It is common to have very long TTL values on a DNS entry and that defines how long a system will cache that entry. Also not everyone will respect low TTL values, and cache for longer anyway.

-Otanx

Dieselboy