Rancher + rke2 limit support of vSphere

1

Hi everyone,
We are currently using rancher and rke1 to create k8s clusters in on-prem infrastructure based on vSphere. Everything works great and thanks suse team for this solution.
Since SUSE announced the end of support for rke1, we started testing the creation of rke2 clusters in rancher on vSphere, but we discovered a new behavior of rancher:

  1. In the cluster creation UI, now Cloud Credentials is a global parameter for the downstream cluster, not Machine Pools, so if we have 2 separate vSphere datacenters, we cannot create nodes/masters in two DCs at the same time, which is not a fault-tolerant scheme.
  2. In the cluster creation UI, there is now no option to select a vSphere Datastore Cluster (with RDS), it is suggested to select a specific Datastore. This means that all k8s cluster nodes(VMs) will be created disk in one Datastore, which is also not a fault-tolerant solution.

Perhaps there are rancher developers here who can comment on whether these problems are planned to be fixed in the future.

2

To achieve this functionality, you can use the cluster template mechanism. For over three years, we have been successfully provisioning clusters on separate vCenters supporting separate DCs.

You can find more information here:

A more specific example of the solution here:

Cheers,
R2

3

Hi @R2D2,
Thanks for your answer.
Currently we use terraform to create clusters, which communicates with the Rancher API - this also works great, i.e. the Rancher API supports this functionality, but the Rancher UI does not and has a lot of alerts and error messages when such a cluster is created.
So it would be interesting to know if the Rancher team plans to fix this in the next releases?

1 Like