@5ubj3ctz3r0 wrote:
@EA_Leeuw
It might be worth looking into the following server on your end: pub-bf-prod-grpc.ops.dice.se
I did a packet capture with, and without VPN. I noticed that without VPN (routing through AT&T directly) there was a ton of TCP retransmissions/TLS errors towards this IP 52.208.207.227. This was not seen when connected to VPN. Could be a red herring, but one of the only things that easily stood out when comparing captures.
Also, a trace route showed no common hops.
Coming from 68.248.0.0/13 when it doesn't work (AT&T)
Coming from 154.47.25.0/24 when it does work (VPN)
@EA_Leeuw
This has to be the issue.
pub-bf-prod-grpc.ops.dice.se resolves to:
108.128.42.217
34.248.44.177
52.208.207.227
When it starts "suddenly working" at night (for me around 10PM-12AM CT) there is no more attempts to connect to 52.208.207.227, it now connects to 108.128.42.217 or 34.248.44.177 (and the game connects successfully). Assuming each of these IPs are a load balancer there is clearly an issue with AT&T sourced IPs connecting to 52.208.207.227.
Looks like you're using Akamai for your DNS so I assume there is some GTM load balancing policies here - something marks 108.128.42.217 as the IP to serve instead of 52.208.207.227. I don't know if all of these are "live" at all times but if this is still going on tomorrow I will try to add a static hosts file entry to mapping pub-bf-prod-grpc.ops.dice.se to 108.128.42.217 and see if it works during the day without use of a VPN. This is assuming that tomorrow it will be serving 52.208.207.227, and that these are active/active setups.
Or this might not be DNS related at all and it's more about how the cloud instances are load balanced, not sure if the application targets one of those IPs based off DNS resolution alone.