Kubecost exporter for Flexera CCO

Kubecost Flexera Exporter is a utility to collect cost allocation data. It is a command line tool that automates the transfer of Kubernetes cluster cost allocation data to Cloud Cost Optimization. This tool generates a CSV file for each day of the current (and optionally previous) month in a format compatible with the Flexera One platform and then uploads files into Cloud Cost Optimization via CBI connect. Kubecost Flexera Exporter utilizes Kubecost Allocation API to request cost allocation data. The majority of Kubecost Allocation API parameters are exposed as exporter settings, matching Kubecost API parameters are listed in the exporter setting descriptions.

Supported Kubecost Allocation API Versions

This exporter is able to work with 1.x, 2.2.3 or higher versions.

With versions from 2.0 to 2.2.2 there could be some unexpected results when using the Allocation API, especially using the shareIdle parameter which in those versions is not correctly implemented. That is why it is not recommended to use any of these versions.

NOTE: For versions 2.2.3 and higher, the properties received from the Kubecost Allocation API depend on the level of aggregation being used, smaller levels of granularity will include higher levels but not vice versa. For instance if you set aggregation as pod, exporter will include data for cluster, namespace, controller, controllerKind, node and pod but if you set aggregation as namespace, exporter won't include data for controller, controllerKind, node and pod.

OpenCost Support

The Kubecost Flexera Exporter now also supports the OpenCost API, which is largely compatible with the Kubecost Allocation API. This means you can easily switch between Kubecost and OpenCost for retrieving Kubernetes cost allocation data, depending on your preference or requirements.

Key Differences

While OpenCost Allocation API mirrors that of Kubecost's for the most part, however, there are a few key differences to be aware of:

Unsupported Parameters: OpenCost does not support the following parameters:
- idleByNode
- shareIdle
- shareNamespaces
- shareTenancyCosts

These parameters are specific to Kubecost's approach to handling idle costs, shared namespace costs, and tenancy costs allocation. If your use case relies on these features, you might need to adjust your cost analysis strategy when using OpenCost. More information can be found in the comparison table.

Default Values for Certain Parameters: When using OpenCost, it's noteworthy that certain parameters have default values distinct from used in Kubecost. Specifically:
- host should be set to opencost.opencost.svc.cluster.local:9003
- apiPath should default to /

Ensure that these values are correctly set in your configuration files to avoid connectivity or functionality issues with the OpenCost service.

Installation

The application can be installed from golang sources, from a docker image or via the helm package manager.

Go sources

This app requires Go version 1.21 or higher. To install, run:

go install github.com/flexera-public/cbi-oi-kubecost-exporter

Settings

The app is configured using environment variables defined in a .env file. The following configuration options are available:

Environment Variable	Description
FILE_PATH	The path where the generated CSV files are stored. Default is "/var/kubecost"
FILE_ROTATION	Indicates whether to delete files generated for previous months. Default is true. Note: current and previous months data is kept.
BILL_CONNECT_ID	The ID of the bill connect to which to upload the data. Default value is "cbi-oi-kubecost-1". To learn more about Bill Connect, and how to obtain your BILL_CONNECT_ID, please refer to Creating Kubecost CBI Bill Connect in the Flexera documentation.
ORG_ID	The ID of your Flexera One organization, please refer to Organization ID Unique Identifier in the Flexera documentation.
REFRESH_TOKEN	The refresh token used to obtain an access token for the Flexera One API. Please refer to Generating a Refresh Token in the Flexera documentation.
SERVICE_APP_CLIENT_ID	The service account client ID used to obtain an access token for the Flexera One API. Please refer to Using a Service Account in the Flexera documentation. This parameter is incompatible with REFRESH_TOKEN, use only one of them.
SERVICE_APP_CLIENT_SECRET	The service account client secret used to obtain an access token for the Flexera One API. Please refer to Using a Service Account in the Flexera documentation.
SHARD	The zone of your Flexera One account. Valid values are NAM, EU or AU.
INCLUDE_PREVIOUS_MONTH	Indicates whether to collect and export previous month. Default is false. This parameter is incompatible with REFRESH_TOKEN, use only one of them.
REQUEST_TIMEOUT	Indicates the timeout per each request in minutes.
KUBECOST_HOST	The hostname of the Kubecost instance. Default is "kubecost-cost-analyzer.kubecost.svc.cluster.local:9090".
KUBECOST_API_PATH	The base path for the Kubecost API endpoint. Default is "/model/"
AGGREGATION	The level of granularity to use when aggregating the cost data. Valid values are namespace, controller, node or pod. Default is pod. Note: Exporter collects namespace labels regardless of set aggregation level and includes them into entity labels.
IDLE	Indicates whether to include cost of idle resources. Valid values are true and false. Default is true.
IDLE_BY_NODE	Indicates whether idle allocations are created on a per node basis. Valid values are true and false. Default is false.
SHARE_IDLE	Indicates whether allocate idle cost proportionally across non-idle resources. Default is false.
SHARE_NAMESPACES	Comma-separated list of namespaces to share costs with the remaining non-idle, unshared allocations. Default = kube-system,cadvisor
SHARE_TENANCY_COSTS	Indicates whether to share the cost of cluster overhead assets across tenants of those resources. Default is true.
MULTIPLIER	Optional multiplier for costs. Default is 1.
CREATE_BILL_CONNECT_IF_NOT_EXIST	Flag to enable automatic creation of Bill Connect. Default is false.
VENDOR_NAME	Vendor name for the Bill Connect. It is used when CREATE_BILL_CONNECT_IF_NOT_EXIST is set to true . Default value is "Kubecost".

Execution

To use this app, run:

flexera-kubecost-exporter

Kubecost exporter helm chart for Kubernetes

There are two different ways to transfer custom Helm configuration values to the kubecost-exporter:

1. Pass exact parameters via --set command-line flags:

helm install kubecost-exporter cbi-oi-kubecost-exporter \
    --repo https://flexera-public.github.io/cbi-oi-kubecost-exporter/ \
    --namespace kubecost-exporter --create-namespace \
    --set flexera.refreshToken="Ek-aGVsbUBrdWJlY29zdC5jb20..." \
    --set flexera.orgId="1105" \
    --set flexera.createBillConnectIfNotExist="true" \
    --set flexera.vendorName="test-vendor" \
    --set flexera.billConnectId="cbi-oi-kubecost-..." \
    ...

2. Pass exact parameters via custom values.yaml file:

2.1 Create a values.yaml file and add the necessary settings to it as below:

flexera:
    refreshToken: 'Ek-aGVsbUBrdWJlY29zdC5jb20...'
    orgId: '1105'
    billConnectId: 'cbi-oi-kubecost-...'
    createBillConnectIfNotExist: 'true'
    vendorName: 'test-vendor'

kubecost:
    aggregation: 'pod'

2.2 Apply this file when installing kubecost-exporter:

helm install kubecost-exporter cbi-oi-kubecost-exporter \
    --repo https://flexera-public.github.io/cbi-oi-kubecost-exporter/ \
    --namespace kubecost-exporter --create-namespace \
    --values values.yaml

Verifying configuration

After successfully installing the helm chart, you can trigger the CronJob manually to ensure that everything is working as expected:

Check the schedule of the CronJob:

kubectl get cronjobs -n <your-namespace>

The SCHEDULE column should reflect the schedule you have set (default: "0 */24 * * *"). The NAME column shows the name of your CronJob.

Manually create a job from the CronJob:

kubectl create job --from=cronjob/<your-cronjob-name> manual-001 -n <your-namespace>

Replace <your-cronjob-name> with the name of your CronJob you obtained from the previous command.

Monitor the logs of the job:

First, get the name of the pod associated with the job you have just created:

kubectl get pods -n <your-namespace>

Look for the pod that starts with the name "manual-001".

Then, fetch the logs:

kubectl logs <your-pod-name> -n <your-namespace>

Replace <your-pod-name> with the name of your pod.

You should see 200/201s in the logs, which indicates that the exporter is working as expected. This means that the CronJob will also run successfully according to its schedule.

Helm configuration Values

Key	Type	Default	Description
activeDeadlineSeconds	int	`10800`	The maximum duration in seconds for the cron job to complete
cronSchedule	string	`"0 /24 * *"`	Setting up a cronJob scheduler to run an export task at the desired time.
env	object	`{}`	Pod environment variables. Example using envs to use proxy: {"NO_PROXY": ".svc,.cluster.local", "HTTP_PROXY": "http://proxy.example.com:80", "HTTPS_PROXY": "http://proxy.example.com:80"}
filePath	string	`"/var/kubecost"`	File path to mount persistent volume.
fileRotation	bool	`true`	Indicates whether to delete files generated for previous months. Note: current and previous months data is kept.
flexera.billConnectId	string	`"cbi-oi-kubecost-1"`	The ID of the bill connect to which to upload the data. To learn more about Bill Connect, and how to obtain your BILL_CONNECT_ID, please refer to Creating Kubecost CBI Bill Connect in the Flexera documentation.
flexera.createBillConnectIfNotExist	string	`"false"`	Flag to enable automatic creation of Bill Connect.
flexera.orgId	string	`""`	The ID of your Flexera One organization, please refer to Organization ID Unique Identifier in the Flexera documentation.
flexera.refreshToken	string	`""`	The refresh token used to obtain an access token for the Flexera One API. Please refer to Generating a Refresh Token in the Flexera documentation. You can provide the refresh token in two ways: 1. Directly as a string: refreshToken: "your_token_here" 2. Reference it from a Kubernetes secret: refreshToken: valueFrom: secretKeyRef: name: flexera-secrets # Name of the Kubernetes secret key: refresh_token # Key in the secret containing the refresh token
flexera.serviceAppClientId	string	`""`	The service account client ID used to obtain an access token for the Flexera One API. Please refer to Using a Service Account in the Flexera documentation. This parameter is incompatible with refreshToken, use only one of them.
flexera.serviceAppClientSecret	string	`""`	The service account client secret used to obtain an access token for the Flexera One API. Please refer to Using a Service Account in the Flexera documentation. This parameter is incompatible with refreshToken, use only one of them.
flexera.shard	string	`"NAM"`	The zone of your Flexera One account. Valid values are NAM, EU or AU.
flexera.vendorName	string	`"Kubecost"`	Vendor name for the Bill Connect. It is used when CREATE_BILL_CONNECT_IF_NOT_EXIST is set to true.
image.pullPolicy	string	`"Always"`
image.repository	string	`"public.ecr.aws/flexera/cbi-oi-kubecost-exporter"`
image.tag	string	`"1.17"`
imagePullSecrets	list	`[]`
includePreviousMonth	bool	`false`	Indicates whether to collect and export previous month.
kubecost.aggregation	string	`"pod"`	The level of granularity to use when aggregating the cost data. Valid values are namespace, controller, node, or pod.
kubecost.apiPath	string	`"/model/"`	The base path for the Kubecost API endpoint.
kubecost.host	string	`"kubecost-cost-analyzer.kubecost.svc.cluster.local:9090"`	Default kubecost-cost-analyzer service host on the current cluster. For current cluster is serviceName.namespaceName.svc.cluster.local
kubecost.idle	bool	`true`	Indicates whether to include cost of idle resources.
kubecost.idleByNode	bool	`false`	Indicates whether idle allocations are created on a per node basis.
kubecost.multiplier	float	`1`	Optional multiplier for costs.
kubecost.shareIdle	bool	`false`	Indicates whether allocate idle cost proportionally across non-idle resources.
kubecost.shareNamespaces	string	`"kube-system,cadvisor"`	Comma-separated list of namespaces to share costs with the remaining non-idle, unshared allocations.
kubecost.shareTenancyCosts	bool	`true`	Indicates whether to share the cost of cluster overhead assets across tenants of those resources.
persistentVolume.enabled	bool	`true`	Enable Persistent Volume. Recommended setting is true to prevent loss of historical data.
persistentVolume.size	string	`"1Gi"`	Persistent Volume size.
requestTimeout	int	`5`	Indicates the timeout per each request in minutes.

Kubecost/Opencost Integration Configuration

Below are the parameters used in the Helm configuration along with the environment variables that are directly employed for API requests.

Helm Value	Environment Variable	Kubecost API Parameter	OpenCost API Parameter	Description
kubecost.aggregation	AGGREGATION	`aggregate`	`aggregate`	Determines the level of granularity to use when aggregating the cost data. Valid values are namespace, controller, or pod.
kubecost.idle	IDLE	`idle`	`includeIdle`	Indicates whether to include the cost of idle resources in the analysis.
kubecost.idleByNode	IDLE_BY_NODE	`idleByNode`	-	Specifies whether idle allocations are created on a per-node basis.
kubecost.shareIdle	SHARE_IDLE	`shareIdle`	-	Indicates whether to allocate idle cost proportionally across non-idle resources.
kubecost.shareNamespaces	SHARE_NAMESPACES	`shareNamespaces`	-	Specifies a comma-separated list of namespaces to share costs with the remaining non-idle, unshared allocations.
kubecost.shareTenancyCosts	SHARE_TENANCY_COSTS	`shareTenancyCosts`	-	Indicates whether to share the cost of cluster overhead assets across tenants of those resources.

License

This tool is licensed under the MIT license. See the LICENSE file for more details.