This topic provides an overview of the Magalix pod aka Agent - 45 seconds read
Magalix Agent is a pod that connects your Kubernetes cluster to Magalix backend. Installing the agent is a very simple process, and once installed the process to analyze your system kicks-in. With the right permissions, it is a single Kubectl command to install Magalix agent.
Running in Autopilot Mode
Magalix agent runs in read-only mode until you enable the Autopilot. In other words, it does not execute any decisions or recommendations. During the read-only mode, it just collects metrics and all the magic takes place in Magalix backend.
Data that Magalix Agent Collects
Magalix Agent only collects necessary data relevant to the autoscaling process. it does not collect any application specific data or have access to any of your container’s volumes. It relies on Kubernetes APIs to collect the following information:
- Nodes description, which is equivalent to kubectl describe nodes
- Pods description, which is equivalent to kubectl describe pods
- Available metrics definitions and values. This one, in particular, is a bit more complex given the different tools that exist right now to monitor Kubernetes. Magalix agent relies on the use of Kubelet APIs and CAdvisor to pull the basic metrics.
Running the Autopilot Mode
In Autopilot mode Magalix agent executes decisions inside the managed cluster. The default mode when the agent is installed is the read-only mode, which means that no decisions will be carried out. You will need to explicitly enable the Autopilot for all your cluster or some namespaces for the agent to execute these decisions.
Autopilot mode is part of the Pro plan. You will need to upgrade your plan before you can enable the autopilot mode.