Kubernetes View version_11.5.01
The Kubernetes view enables you to manage the capacity and efficiency of containerized environments managed by the Kubernetes platform, and presents key capacity metrics and charts for Kubernetes clusters, nodes, namespaces, deployments and pods.
You can use the Kubernetes view to complete tasks such as:
- Understand resource bottlenecks and aggregate residual capacity of Kubernetes clusters as well as individual nodes
- Detect current or imminent resource saturation conditions and days before the resource is saturated for every major Kubernetes resource (e.g. cluster, nodes, deployments, pods)
- Assess the level of infrastructure efficiency, by comparing allocated vs actually resources, and identify most wasteful deployments or pods
- Identify application in resource usage patterns and detect resource shortage conditions
- Characterize the footprint of infrastructure resources on a containers' image basis
- Understand resource utilization of namespaces, including the level of usage of resource quotas
Videos
The following video (5:35) provides a brief introduction of the Kubernetes views.
The following video (9.53) provides information about how to access and use the Kubernetes views.
Requirements
Supported versions of TrueSight Capacity Optimization
- Supported TrueSight Capacity Optimization versions: 10.7.01 onward
Certified data sources
The certified data sources for the views are the following:
- Kubernetes (K8S) Heapster
- Kubernetes (K8S) Prometheus
You must configure and run one of these ETL modules to have data visible in the view
Data Visibility
Conventions
The Kubernetes view provides summarized, high-level capacity KPIs designed for capacity management.The following common conventions around naming and metrics aggregation are valid for all data presented in the view:
- Standard table column names: Metric name [unit of measurement] (ex. "Memory [GB]")
- Unit of measurement is omitted when implicitly evident
- The metric value is the aggregation of the last 30 days
- The metric value is computed as follows: for each day, the daily peak is considered (at hourly resolution). Then, the mean value of the daily peak over the last 30 days is shown.
This is only valid for summary metrics presented in tables and overall page. The charts presented in the details pages follow regular over time metrics semantics, whose time frame and time resolution available as filters in the top of the page.
View Structure
The Kubernetes view is composed of the following first-level pages:
- Overview: it presents a summary of Kubernetes components states
- Clusters: it shows capacity metrics for Kubernetes clusters
- Nodes: it shows capacity metrics for Kubernetes nodes
- Namespaces: it shows capacity metrics for Kubernetes namespaces
- Deployments: it shows capacity metrics for Kubernetes deployments
- Pods: it shows capacity metrics for Kubernetes pods
For each page except Overview, a set of second-level tabs cover information about some or all of the following parts:
- Capacity: summary of the most important capacity indicators
- CPU: the most relevant CPU configuration and performance metrics
- Memory: the most relevant memory configuration and performance metrics
- Storage: the most relevant storage configuration and performance metrics (this tab is defined only in the Cluster page)
From all of the second-level tabs it is possible to drill-down to an entity detail page, which presents the most relevant performance metrics as time charts and tables for the most important configuration properties.
Thresholds, Status & Bottleneck
The table below reports the thresholds used in the view.
The column Resource is defined as: "Resource taken into account: status"; the defined resources are:
- CPU: all the metrics related to the CPU (ex. CPU utilization or CPU Request over CPU Number)
- MEM: all the metrics related to the memory (ex. memory utilization or memory Request over memory capacity)
- SAT: days before saturation of the resource
- POD: percentage of pods required over pod creation limit (this limit depends from the Kubernetes version installed)
The status is evaluated as the worst status among the following resources and the bottleneck is the worst resource associated to that state.
Please note CPU/Memory Used on Request metrics are not considered in the status and bottleneck evaluation since used is allowed to be greater than requested, hence they can go above 100%.
Materializer Task
The Kubernetes view takes advantage of TrueSight Capacity Optimization materialized data marts in order to enable faster page loading times during user browsing, by pre-computing the underlying data. To achieve this result, a proper DataMartMaterializer task is deployed as part of the view installation process. By default, the task is scheduled to run every day at midnight. Please modify the scheduling parameters to fit the environment needs, such as:
- Data loading scheduling and warehouse latency
- Multiple refresh in the same day
- Kubernetes View user need
The Kubernetes View materializer is shown in the next figure and can be found under System Tasks.