如何为 Pod 之间的平衡请求设置 kube-proxy？

Question

I have a k8s cluster with 2 nodes. I have a ClusterIP Service with a Deployment linked to a HPA that autoscale a group of Pods. I would like to send many requests in multiple threads to the service and balance the requests in the current number of Pods available in that moment. At starting i was using the default kube-proxy mode IPTABLES but was RANDOM assigning the requests to Pods with some Pods working more and others less, increasing the response time of requests and having a not high percentage in average of CPU for add new pods. Than I switched to IPVS with "sed" scheduler but was not balancing the same because every thread was assigned to a Pod and working at max while last created Pods where not working. Could you give me some hints of how can I reach my goal?

Answer 1

Kube-proxy does not support virtual IPs as a destination Every node in a Kubernetes cluster runs a kube-proxy. kube-proxy is responsible for implementing a form of virtual IP for Services of type other than ExternalName kube-proxy can run in three different modes:

• iptables (default mode)
• Ipvs
• userspace (“legacy” mode, not recommended anymore) While the iptables mode is totally fine for many clusters and workloads, ipvs can be useful when the number of services is important (more than 1,000). Indeed, as iptables rules are read sequentially, its usage can impact the routing performances if many services exist in the cluster. You can look at thisdocument for the configuration.