CAST AI Components Hosted On Customers' Clusters

Connecting existing cluster procedure installs several CAST AI components into a cluster.
This is done in phases to provide different levels of experience:

Phase 1 - is meant to provide visibility around connected clusters without possibility of tuning it; one can think of Phase 1 as about operating in read-only mode.
Phase 2 - enables all the functionality of CAST AI platform mostly around clusters optimisation; CAST AI platform instructs clusters and Cloud Providers to re-arrange used resources to reach most optimal state.

Phase 1 Component - CAST AI Kubernetes Agent

CAST AI Agent is the first component installed when a new cluster is connected.
The agent runs as a Pod in a CAST AI dedicated namespace:

» kubectl get pods -n castai-agent
NAME                                         READY   STATUS    RESTARTS   AGE
castai-agent-7f9d7ff65b-8qm7p                1/1     Running   0          78m
castai-agent-cpvpa-56f749fb-n2wzp            1/1     Running   0          22d

CAST AI Kubernetes Agent is responsible for sending cluster state data (snapshots) to the main system
Cluster Proportional Vertical Autoscaler is responsible for tuning allocated resource for castai-agent Pods based on predefined formula

Phase 2 Autoscaling Components

CAST AI Cluster Controller, Evictor and Spot Handler (installed as DaemonSet, not as a regular Deployment) components are installed when a connected cluster is promoted to Phase 2, which enables cost savings by managing customer's cluster:

❯ k get pods -n castai-agent
NAME                                             READY   STATUS    RESTARTS   AGE
castai-agent-7f9d7ff65b-8qm7p                    1/1     Running   0          80m
castai-agent-7f9d7ff65b-kf2zp                    1/1     Running   0          5h7m
castai-agent-cpvpa-56f749fb-n2wzp                1/1     Running   0          22d
castai-cluster-controller-757997ff6c-r6x25       1/1     Running   0          27d
castai-cluster-controller-757997ff6c-xw54g       1/1     Running   0          27d
castai-evictor-5684748495-kl2q4                  1/1     Running   0          22d
castai-kvisor-787c5dd946-gmzs5                   1/1     Running   0          6d18h
castai-spot-handler-44shj                        1/1     Running   0          43m

Cluster controller is responsible for executing actions it receives from the central platform (like for example accept a newly created node to the cluster, etc.)
Evictor is responsible for removing pods from underutilised nodes to be able to decrease overall amount of cluster nodes
Spot Handler is responsible for scheduled events monitoring (provided by Instance Metadata Service) and delivering them to the central platform

Phase 2 Security Components - Kvisor

kvisor is resposible for images vulnerability scanning, Kubernetes YAML manifests linting and CIS security recommendations.