Skip to content

Start Observing

AI platform enables the management and creation of multicloud and multiple clusters. Building upon this capability, Insight serves as a unified observability solution for multiple clusters. It collects observability data from multiple clusters by deploying the insight-agent plugin and allows querying of metrics, logs, and trace data through the AI platform Insight.

insight-agent is a tool that facilitates the collection of observability data from multiple clusters. Once installed, it automatically collects metrics, logs, and trace data without any modifications.

Clusters created through Container Management come pre-installed with insight-agent. Hence, this guide specifically provides instructions on enabling observability for integrated clusters.

As a unified observability platform for multiple clusters, Insight's resource consumption of certain components is closely related to the data of cluster creation and the number of integrated clusters. When installing insight-agent, it is necessary to adjust the resources of the proper components based on the cluster size.

  1. Adjust the CPU and memory resources of the Prometheus collection component in insight-agent according to the size of the cluster created or integrated. Refer to Prometheus resource planning.

  2. As the metric data from multiple clusters is stored centrally, AI platform administrators need to adjust the disk space of vmstorage based on the cluster size. Refer to vmstorage disk capacity planning.

  3. For instructions on adjusting the disk space of vmstorage, refer to Expanding vmstorage disk.

Since AI platform supports the management of multicloud and multiple clusters, insight-agent has undergone partial verification. However, there are known conflicts with monitoring components when installing insight-agent in Suanova 4.0 clusters and Openshift 4.x clusters. If you encounter similar issues, refer to the following documents:

Currently, the insight-agent collection component has undergone functional testing for popular versions of Kubernetes. Refer to: