What is the difference between the IPVS/IPTABLES mapping rules and the Endpoint object that has a list of pod ips for a service ?
Are they not serving the same purpose of resolving a request to a service to a backend pod ?
At which point in a request flow do these apply ?
Some details about these two please.
1 Answers
IPVS/IPTABLES are (load-balancer/firewall feature) modes available on Kube-proxy running on each node in a cluster while Endpoint object is a component that works hand-in-hand with the Service object in a cluster base on Label. IPVS/IPTABLES help pod route request out the node is running on to pod on another node while Endpoint helps the Service object to route accepted request to the destination pod pool with the same Label as Service object. Yes @Aftab, both serve as a routing component in the cluster but at different levels.
This is exactly what I wanted to know. Would it be fair to say that IPVS/IPTABLES acts like a L2 switch (the cbr0 bridge) while Endpoint object acts like a router?