Prerequisites
Introduction
This topic describes the prerequisites for installing KubeSlice. Ensure that you have the following environment set up that is required to install KubeSlice. In this environment, you must install the KubeSlice Controller and register clusters with it. You need:
- Kubernetes clusters.
- Clusters to be able to reach each other (that is, for each cluster, there must be an IP address that is routable from other clusters that want to reach it) whether through a public or a private network.
- Ports exposed on the clusters so that the tunneling service can connect.
- A kubeconfig file retrieved from the clusters for authentication.
Command Line Tools
You need the following command line tools to install the KubeSlice.
| Package Required | Installation Instructions |
|---|---|
| Helm - The Package Manager for Kubernetes | Installing Helm |
| kubectl - Kubernetes CLI | Installing Tools |
| kubectx and kubens | Installing Kubectx |
| Docker | https://docs.docker.com/engine/install/ |
| Kind CLI | https://kind.sigs.k8s.io/docs/user/quick-start/#installation (for kind clusters) |
Infrastructure Requirements
Kind Clusters
The following are the infrastructure requirements to install KubeSlice components.
If you are on Ubuntu OS, then it is recommended to increase the ulimit
to 2048 or unlimited.
If you still face issues, see errors due to too many open files.
| Host Requirements | Minimum of 8vCPUs and 8GB of RAM |
Ensure to modify the memory and CPU usage allowed to docker as described in
https://docs.docker.com/desktop/windows/#resources.
Requirements for Hosting KubeSlice Controller
You can install the KubeSlice Controller on a cluster and also register the same cluster with the KubeSlice Controller.
| Cluster Requirements | 1 Kubernetes Cluster |
| Supported Kubernetes Versions | 1.23 and 1.24 |
| Required Helm Version | 3.7.0 |
Requirements for Worker Clusters
| Minimum Clusters Required | 2 Kubernetes Clusters |
| Nodes Reserved for KubeSlice Components | 1 Node per cluster |
| Supported Kubernetes Versions | 1.23 and 1.24 |
| Required Helm Version | 3.7.0 |
Cloud Clusters
The following are the infrastructure requirements to install KubeSlice components on cloud clusters.
Requirements for Hosting KubeSlice Controller
You can install the KubeSlice Controller on a cluster and also register the same cluster with the KubeSlice Controller.
| Cluster Requirements | 1 Kubernetes Cluster |
| Supported Kubernetes Versions | 1.23 and 1.24 |
| Supported Kubernetes Services | Azure Kubernetes Service (AKS), Elastic Kubernetes Service (EKS), and Google Kubernetes Engine (GKE) |
| Required Helm Version | 3.7.0 |
Kubeslice has been tested on AKS, GKE, and EKS, but it should run in any Kubernetes environment with the supported Kubernetes versions.
Requirements for Worker Clusters
| Minimum Clusters Required | 1 Kubernetes Clusters |
| Minimum Nodes or NodePools Required | 2, Each with a minimum of 4 vCPUs and 16 GB of RAM |
| NodePools Reserved for KubeSlice Components | 1 NodePool |
| Cluster Requirements | - 1 Kubernetes Cluster - 2 minimum nodes or NodePools each with a minimum of 4 vCPUs and 16 GB of RAM. - 1 NodePool for KubeSlice components |
| Supported Kubernetes Versions | 1.23 and 1.24 |
| Supported Kubernetes Services | Azure Kubernetes Service (AKS), Elastic Kubernetes Service (EKS), and Google Kubernetes Engine (GKE) |
| Required Helm Version | 3.7.0 |
| Required Istio Version | 1.13.3 |
Cluster Networking
Open the UDP Ports required for the KubeSlice Gateway Node in both public and private clusters.
| Kubernetes Service | CNI with Network Policy Support | Open UDP Ports | Reachability Options |
|---|---|---|---|
| Azure Kubernetes Service | Enable Calico for Azure AKS | 30000-33000 | Public Node IP with NodePort Service |
| Elastic Kubernetes Service | Enable Calico for Amazon EKS | 30000-33000 | Public Node IP with NodePort Service |
| Google Kubernetes Engine | Enable Calico for Google GKE | 30000-33000 | Public Node IP with NodePort Service |
Required Cluster Roles (RBAC)
Only the cluster administrator can install the KubeSlice Controller and the Slice Operator (Worker Operator).
Cluster Context Switching
You must change your kubeconfig context frequently to run kubectl commands in each cluster. Using kubectx can make this
process easier and save you time.
To switch the context to a certain cluster, run the following command with the corresponding cluster name:
kubectx <cluster name>
Expected Output.
Switched to context "<cluster name>"
To return to the last used context, use the following command:
kubectx -
Expected Output.
Switched to context "<previous context>"
You have successfully set up the required KubeSlice prerequisites.
