Proxied Round Robin DNS only using one server

I have the following Round Robin DNS setup currently:

My problem is the following:

  • If I turn off one server, everything works
  • If I turn off the other server, everything fails

I also tested it with both servers working and loading many small requests. Using nload, one server gets all the traffic, the other gets nothing.

Why is this and how can I possibly debug this?


Here is what I’m tried for debugging so far:

nslookup tiles.openfreemap.org

Name:	tiles.openfreemap.org
Address: 172.67.190.138
Name:	tiles.openfreemap.org
Address: 104.21.92.85

Now, I try the following:

curl --resolve tiles.openfreemap.org:443:104.21.92.85 -I https://tiles.openfreemap.org/planet
and
curl --resolve tiles.openfreemap.org:443:172.67.190.138 -I https://tiles.openfreemap.org/planet

Both of them report HTTP/2 521, if the first server is down. What happens with the second one, doesn’t matter.

I tested it in production and both origin servers receive kind of equal amounts of traffic.

So there is some kind of smart routing happening even without Load Balancer enabled. I suspect for every visitor one origin server is selected and then it’s “sticky”: all connections go to that one server, no matter what happens. For a different visitor the other server might be selected in a “sticky” way.

This topic was automatically closed 15 days after the last reply. New replies are no longer allowed.