Kong results in 504 errors when nodes scale in.
Is there a way to update keepalive_time and keepalive_timeout for the nginx kong uses?
Are there any other settings I should look at?
The architecture is as follows:
ALB → NLB → EKS NodePort (kong)
Managed EKS nodes are not draining gracefully, presumably because the NLB is not terminating the connection to kong, which results in 504 errors on the ALB.