I have a bunch of pods and corresponding cluster IP services running in a Kubenetes cluster. I use an Ingress controller to route incoming traffic to these workloads.
Now, when I make an external request from within a workload, I would like those requests to go out from a static IP.
Are both incoming and outgoing traffic handled via the same network interface?
How do I set a static IP for my outgoing requests from a pod
Thanks in advance.
Some research showed me the following method and it worked:
Create a new node pool with a specific label (lets say ip=static)
Deploy the workload with the following selector
nodeSelector:
ip: staticThrough VPC Network, reserve an external static IP
Assign this IP to the VM which is used by the new node pool
However, as the @user578582 has mentioned, when the scaling is done, I will have to use a NAT gateway to make all the outbound traffic to go from a static IP under this configuration.
If you're on AWS or GKE the only way to get a static outbound from an (autoscaling) node pool is via a NAT gateway. On GKE you are also required to have private nodes (no public ips).
Kubernetes ingress, at least in GKE, is treated separately from egress and is a different load balancer.