Rancher 2.0 cluster K8S deployment strategy RKE or Custom


I plan to se rancher 2.0 and create a K8S cluster (on AWS)

I would like to know the difference between two way:

1 .First solution

-> Deploy is possible with custom cluster from rancher, we just have executed the docker run command provide by rancher on workers/etcd/masters.

  1. Second solution

-> Use RKE to build a K8S cluster and import from Rancher

At this time, i don’t see really the differences, i can’t find article about that … What is your opinion ?

The last question, we can use K8S official image and Rancher K8S image (a fork i imagine). What is the real difference ? (no documentation)

Many thanks folks !

I’m newbie in k8s, but i saw that there many solutions too deploy it.



I saw 2 approach To deploy k8s HA for production. The most important is to be able to maintain the HA config.

  • one solution is to use k8s deployement functionality provided by the on premise or IaaS provider. Kind of “All inclusive” with Ubuntu (maas/openstack/juju) or vmware or redhat openshit, or coreos tectonic, etc …

But this approche is too depending or the provider.

  • second approach is to focus only one k8s deployement for HA based on existing node/VM ever deployed.
    I tested RKE and kubeadm
    But kubeadm la still in draft for HA deployement
    There is also kubespray or kops but i didn’t test it