How to resolve Kubernetes DNS troubles with Amazon RDS hosts

Question

I have an Flask + SQLAlchemy deployed in Kubernetes, using Amazon AWS as cloud provider (Amazon EC2 machines) with a PostgreSQL RDS.

Sometimes an excetion arises when a request is performed:

psycopg2.OperationalError: could not translate host name "xxx.yyy.eu-west-1.rds.amazonaws.com" to address: Name or service not known

The Amazon RDS instance is located at the same VPC that the Kubernetes cluster, so DNS resolutions must turn into internal IP within the VPC.

I've tried to use Kubernetes kube-dns built-in stubDomains forcing to resolve *.rds.amazonaws.comDNS requests through VPC DNS resolver (in my case 10.0.0.2) but the exception keeps arising.

Does anybody have an idea of what's going wrong?

Thanks in advance.

Answer 1

I have seen that on our flannel backed cluster when the machines are under high CPU load. Also, as a for your consideration, I have never had to alter the kube-dns settings, since it will resolve .cluster.localnames, and then delegate to the EC2 resolv.conf as expected. Have you tried altering the Pod's