I've been migrating a traffic-heavy (https://jsonip.com) service I've run for a number of years to an upgraded server over the last week. Before the dns switch, I was averaging 9mb/s ipv4 inbound. After the switch, its barely 1mb/s on the new server. My inbound ipv6 traffic seems to be unaffected.
I'm trying get some suggestions as to what might be happening.
These graphs represent the before and after.
Old server: https://www.dropbox.com/s/y3wso9y7k6ijvwf/Screenshot%202016-07-31%2016.56.49.png?dl=0
New server: https://www.dropbox.com/s/ahjvhy70vnmlxzf/Screenshot%202016-07-31%2016.57.54.png?dl=0
The first image shows ipv4(top) and ipv6(bottom) traffic for the last 30 days on the old server. The second image is the last 30 days on the new server. You can see where the switchover happened a few days ago. (Notice that the blue line is inbound traffic and green is outbound. The new server is now showing high outbound traffic because I enabled https via LetsEncrypt yesterday and http traffic is being 301'd to https now. So this is expected and not related to the inbound traffic problem.)
When I noticed this happening a couple days ago, I immediately started checking for dns propagation problems. As far as I have been able to tell through multiple global propagation checking sites the rollout has been complete.
I've checked with my host provider and basically got the "looks normal on our end" brushoff.
One pattern I see is that inbound traffic is always much higher than outbound. This is because the inbound traffic contains lots of headers based on the traffic source, and the outbound is always a json object and a few CORS headers that's roughly always the same size. But after the switch over, the graph shows a nearly 1-1 ratio of inbound and outbound traffic. This makes me suspicious that the host's reporting daemon isn't 100% to be trusted.
If anyone more familiar with this part of server networking has some clues as to what's happening, I'd be really grateful.
