Load balancing where ressources are available

2/25/2019

I would like to know how could i setup a load balancer at the kubernetes level so my services points to the pods that has the more ram/cpu ressources available and not randomly.

It would be co-existing with a Horizontal Pod Autoscaler.

As of today i have an ingress pointing to my NodePort service

It would be better if it could work localy with minikube without needing to be on a cloud provider.

If not, im using OVH as my Kubernetes cloud provider.

Thanks for any infos

-- lifeguru42
autoscaling
kubernetes
kubernetes-ingress
load-balancing

1 Answer

3/4/2019

Q: services points to the pods that has the more ram/cpu ressources available

There is no such feature in Kubernetes at the moment.

But you can switch kube-proxy to ipvs mode and it allows you to use the following balancing algorithm:

  • rr: round-robin
  • lc: least connection
  • dh: destination hashing
  • sh: source hashing
  • sed: shortest expected delay
  • nq: never queue

Calico also supports it.

-- VAS
Source: StackOverflow