The website uses cookies to optimize your user experience. Using this website grants us the permission to collect certain information essential to the provision of our services to you, but you may change the cookie settings within your browser any time you wish. Learn more
I agree
Text direction?


In a nutshell, kube-opex-analytics or literally Kubernetes Opex Analytics is a tool to help organizations track the resources being consumed by their Kubernetes clusters to prevent overpaying. To do so it generates, short-, mid- and long-term usage reports showing relevant insights on what amount of resources each project is spending over time. The final goal being to ease cost allocation and capacity planning decisions with factual analytics.

Table of Contents


kube-opex-analytics periodically collects CPU and memory usage metrics from Kubernetes's APIs, processes and consolidates them over various time-aggregation perspectives (hourly, daily, monthly), to produce resource usage reports covering up to a year. The reports focus on namespace level, while a special care is taken to also account and highlight shares of non-allocatable capacities.

Fundamentals Principles

kube-opex-analytics is designed atop the following core concepts and features:

  • Namespace-focused: Means that consolidated resource usage metrics consider individual namespaces as fundamental units for resource sharing. A special care is taken to also account and highlight non-allocatable resources .
  • Hourly Usage & Trends: Like on public clouds, resource consumption for each namespace is consolidated on a hourly-basic. This actually corresponds to the ratio (%) of resource used per namespace during each hour. It's the foundation for cost calculation and also allows to get over time trends about resources being consuming per namespace and also at the Kubernetes cluster scale.
  • Daily and Monthly Usage Costs: Provides for each period (daily/monthly), namespace, and resource type (CPU/memory), consolidated cost computed given one of the following ways: (i) accumulated hourly usage over the period; (ii) actual costs computed based on resource usage and a given hourly billing rate; (iii) normalized ratio of usage per namespace compared against the global cluster usage.
  • Occupation of Nodes by Namespaced Pods: Highlights for each node the share of resources used by active pods labelled by their namespace.
  • Efficient Visualization: For metrics generated, kube-opex-analytics provides dashboards with relevant charts covering as well the last couple of hours than the last 12 months (i.e. year). For this there are built-in charts, a Prometheus Exporter along with Grafana Dashboard that all work out of the box.

Cost Models

Cost allocation models can be set through the startup configuration variable KOA_COST_MODEL. Possible values are:

  • CUMULATIVE_RATIO: (default value) compute costs as cumulative resource usage for each period of time (daily, monthly).
  • RATIO: compute costs as normalized ratios (%) of resource usage during each period of time.
  • CHARGE_BACK: compute actual costs using a given cluster hourly rate and the cumulative resource usage during each period of time.

Read the Configuration section for more details.


Before diving to concepts and technical details in the next sections, the below screenshots illustrate reports leveraged via the kube-opex-analytics's built-in charts or via Grafana backed by the kube-opex-analytics's built-in Prometheus exporter.

Last Week Hourly Resource Usage Trends

Two-weeks Daily CPU and Memory Usage

One-year Monthly CPU and Memory Usage

Nodes' Occupation by Pods

Grafana Dashboard

This is a screenshot of our official one backed by the kube-opex-analytics's built-in Prometheus Exporter.

Getting Started

Kubernetes API Access

kube-opex-analytics needs read-only access to the following Kubernetes APIs.

  • /apis/metrics.k8s.io/v1beta1
  • /api/v1

You need to provide the base URL of the Kubernetes API when starting the program (see example below).

Typically if you're planning an installation inside a Kubernetes cluster, you can connect to the local cluster API endpoint at: https://kubernetes.default.

Likewise, if you're planning an installation outside a Kubernetes cluster you can use a proxied access to Kubernetes API as follows:

$ kubectl proxy

This will open a proxied access to Kubernetes API at

Configuration Variables

These configuration variables shall be set as environment variables before the startup of the service.

kube-opex-analytics supports the following environment variables when it starts:

  • KOA_DB_LOCATION sets the path to use to store internal data. Typically when you consider to set a volume to store those data, you should also take care to set this path to belong to the mounting point.
  • KOA_K8S_API_ENDPOINT sets the endpoint to the Kubernetes API.
  • KOA_COST_MODEL (version >= 0.2.0): sets the model of cost allocation to use. Possible values are: CUMULATIVE_RATIO (default) indicates to compute cost as cumulative resource usage for each period of time (daily, monthly); CHARGE_BACK calculates cost based on a given cluster hourly rate (see KOA_BILLING_HOURLY_RATE); RATIO indicates to compute cost as a normalized percentage of resource usage during each period of time.
  • KOA_BILLING_HOURLY_RATE (required if cost model is CHARGE_BACK): defines a positive floating number corresponding to an estimated hourly rate for the Kubernetes cluster. For example if your cluster cost is $5,000 dollars a month (i.e. ~30*24 hours), its estimated hourly cost would be 6.95 = 5000/(30*24).
  • KOA_BILLING_CURRENCY_SYMBOL (optional, default is '$'): sets a currency string to use to annotate costs on reports.

Deployment on Docker

kube-opex-analytics is released as a Docker image. So you can quickly start an instance of the service by running the following command:

$ docker run -d \
        --net="host" \
        --name 'kube-opex-analytics' \
        -v /var/lib/kube-opex-analytics:/data \
        -e KOA_DB_LOCATION=/data/db \
        -e KOA_K8S_API_ENDPOINT= \

In this command:

  • We provide a local path /var/lib/kube-opex-analytics as data volume for the container. That's where kube-opex-analytics will store its internal analytics data. You can change this local path to another location, but please keep the container volume /data as is.
  • The environment variable KOA_DB_LOCATION points to the container path to store data. You may note that this directory belongs to the data volume atached to the container.
  • The environment variable KOA_K8S_API_ENDPOINT set the address of the Kubernetes API endpoint.

Get Access to the User Interface

Once the container started you can open access the kube-opex-analytics's web interface at http://<DOCKER_HOST>:5483/. Where <DOCKER_HOST> should be replaced by the IP address or the hostmane of the Docker server.

For instance, if you're running Docker on your local machine the interface will be available at:

Due to the time needed to have sufficient data to consolidate, you may need to wait almost a hour to have all charts filled. This is a normal operations of kube-opex-analytics.

Deployment on a Kubernetes cluster

There is a Helm chart to ease the deployment on Kubernetes using, either Helm 2 (i.e with Tiller), Helm 3 (without Tiller) or kubectl.

In each of the cases, check the values.yaml file to customize the configuration options according to your specific requirements.

In particular, you may need to customize the default settings used for the persistent data volume, the Prometheus Operator and its ServiceMonitor, the security context, and many others.

Security Context: kube-opex-analytics's pod is deployed with a unprivileged security context by default. However, if needed, it's possible to launch the pod in privileged mode by setting the Helm configuration value securityContext.enabled to false.

In the next deployment commands, it's assumed that the target namespace kube-opex-analytics exists. You thus need to create it first or, alternatively, adapt the commands to use any other namespace of your choice.

Installation using Helm 2 (i.e. with tiller)

Helm 2 requires to have tiller installed on the cluster.

helm upgrade \
  --namespace kube-opex-analytics \
  --install kube-opex-analytics \

Installation using Helm 3 (i.e. without tiller)

Helm 3 does not longer require to have tiller.

As a consequence the below command shall work with a fresh installation of kube-opex-analytics or a former version installed with Helm 3. There is a known issue when there is already a version not installed with Helm 3.

helm upgrade \
  --namespace kube-opex-analytics \
  --install kube-opex-analytics \

Installation using Kubectl

This approach requires to have the Helm client (version 2 or 3) installed to generate a raw template for kubectl.

$ helm template \
  kube-opex-analytics \
  --namespace kube-opex-analytics \
  helm/kube-opex-analytics/ | kubectl apply -f -

Get Access to UI Service

The Helm deploys an HTTP service named kube-opex-analytics on port 80 in the selected namespace, providing to the built-in dashboard of kube-opex-analytics.

Export Charts and Datasets (PNG, CSV, JSON)

Any chart provided by kube-opex-analytics can be exported, either as PNG image, CSV or JSON data files.

The exportation steps are the following:

  • Get access to kube-opex-analytics's interface.

  • Go to the chart that you want to export dataset.

  • Click on the tricolon icon near the chart title, then select the desired export format.

  • You're done, the last step shall download the result file instantly.

Prometheus Exporter

Starting from version 0.3.0, kube-opex-analytics enables a Prometheus exporter through the endpoint /metrics.

The exporter exposes the following metrics:

  • koa_namespace_hourly_usage exposes for each namespace its current hourly resource usage for both CPU and memory.
  • koa_namespace_daily_usage exposes for each namespace and for the ongoing day, its current resource usage for both CPU and memory.
  • koa_namespace_monthly_usage exposes for each namespace and for the ongoing month, its current resource usage for both CPU and memory.

The Prometheus scraping job can be configured like below (adapt the target URL if needed). A scraping interval less than 5 minutes (i.e. 300s) is useless as kube-opex-analytics would not generate any new metrics in the meantime.

  - job_name: 'kube-opex-analytics'
    scrape_interval: 300s
      - targets: ['kube-opex-analytics:5483']

When the option prometheusOperator is enabled during the deployment (see Helm values.yaml file), you have nothing to do as the scraping should be automatically configured by the deployed Prometheus ServiceMonitor.

Grafana Dashboards

You can either build your own Grafana dashboard or use our official one.

This official Grafana dashboard looks as below and is designed to work out-of-the box with the kube-opex-analytics's Prometheus exporter. It requires to set a Grafana varianle named KOA_DS_PROMETHEUS, which shall point to your Prometheus server data source.

The dashboard currently provides the following reports:

  • Hourly resource usage over time.
  • Current day's ongoing resource usage.
  • Current month's ongoing resource usage.

You should notice those reports are less rich compared against the ones enabled by the built-in kube-opex-analytics dashboard. In particular, the daily and the monthly usage for the different namespaces are not stacked, neither than there are not analytics for past days and months. These limitations are inherent to how Grafana handles timeseries and bar charts.

License & Copyrights

This tool (code and documentation) is licensed under the terms of Apache License 2.0. Read the LICENSE file for more details on the license terms.

The tool includes and is bound to third-party libraries provided with their owns licenses and copyrights. Read the NOTICE file for additional information.

Support & Contributions

We encourage feedback and always make our best to handle any troubles you may encounter when using this tool.

Here is the link to submit issues: https://github.com/rchakode/kube-opex-analytics/issues.

New ideas are welcomed, please open an issue to submit your idea if you have any one.

Contributions are accepted subject that the code and documentation be released under the terms of Apache 2.0 License.

To contribute bug patches or new features, you can use the Github Pull Request model.

Related Notes
Get a free MyMarkup account to save this article and view it later on any device.
Create account

End User License Agreement

Summary | 16 Annotations
track the resources being consumed by their Kubernetes clusters to prevent overpaying
2020/06/10 10:18
2020/06/10 10:18
2020/06/10 10:18
long-term usage
2020/06/10 10:18
ease cost allocation
2020/06/10 10:18
capacity planning decisions
2020/06/10 10:18
hourly, daily, monthly
2020/06/10 10:19
resource usage reports
2020/06/10 10:19
namespace level
2020/06/10 10:19
2020/06/10 10:20
Hourly Usage & Trends
2020/06/10 10:20
Daily and Monthly Usage Costs
2020/06/10 10:20
Occupation of Nodes by Namespaced Pods
2020/06/10 10:20
Efficient Visualization
2020/06/10 10:20
2020/06/10 10:24
2020/06/10 10:24