Skip to content

Latest commit

 

History

History
169 lines (140 loc) · 8 KB

reportdatasources.md

File metadata and controls

169 lines (140 loc) · 8 KB
Raw

Report Data Sources

A ReportDataSource is a custom resource that represents how to store data, such as where it should be stored, and in some cases, how the data is to be collected.

There are currently four types of ReportDataSource's: prometheusMetricsImporter, awsBilling, reportQueryView and prestoTable. Each has a corresponding configuration section within the spec of a ReportDataSource. The main effect that creating a ReportDataSource has is that it causes the metering operator to create a table in Presto or Hive. Depending on the type of ReportDataSource it then may do other additional tasks. For prometheusMetricsImporter datasources the operator periodically collects metrics and stores them in the table. For awsBilling, the operator configures the table to point at an S3 bucket containing AWS Cost and Usage reports, making these reports exposed as a database table. To read more details on how the different ReportDataSources work, read the metering architecture document.

Fields

  • prometheusMetricsImporter: If this section is present, then the ReportDataSource will be configured to periodically poll Prometheus for metrics using the specified Prometheus query.
    • query: The PromQL query to use.
    • storage: This section controls the StorageLocation options, allowing you to control on a per ReportDataSource level, where data is stored.
      • storageLocationName: The name of the StorageLocation resource to use.
    • prometheusConfig:
      • url: If present, the URL of the Prometheus instance to scrape for this ReportDataSource.
  • awsBilling: If specified, the ReportDataSource will be configured to use an S3 bucket containing AWS billing reports as its source of data.
    • source:
      • bucket: Bucket name to store data into.
      • prefix: Path within the bucket where to store data.
      • region: The region where bucket is located.
  • reportQueryView: If this section is present, then the ReportDataSource will be configured to create a View in Presto using the rendered spec.query as the query for the view.
    • queryName: The name of a ReportQuery to create a view from.
    • inputs: Used to override or set values defined in a ReportQuery's spec.input field. For details on how inputs can be specified read the Specifying Inputs section of the ReportQueries documentation.
    • storage: This section controls the StorageLocation options, allowing you to control on a per ReportDataSource level, where data is stored.
      • storageLocationName: The name of the StorageLocation resource to use.
  • prestoTable: If present, then the ReportDataSource will simply make it possible to reference a database table within Presto as a ReportDataSource.
    • tableRef: The name of the PrestoTable that this ReportDataSource should refer to.

PrometheusMetricsImporter Datasource

For ReportDataSources with a spec.prometheusMetricsImporter present, their tables have the following database table schema:

  • timestamp: The type of this column is timestamp. This is the time which the metric was collected.
    • Note: timestamp is also a reserved keyword (for the column type) in Presto, meaning any queries using it must use quotes to refer to the column, like so: SELECT "timestamp" FROM datasource_unready_deployment_replicas LIMIT 1;
  • timeprecision: The type of this column is a double. This is "query resolution step width" used to query this metric from Prometheus. This defines how accurate the data is. The bigger the value, the less accurate. This value is controlled globally by the operator, and has a default value of 60.
  • labels: The type of this column is a map(varchar, varchar). This is the set of Prometheus labels and their values for the metric.
  • amount: The type of this column is a double. Amount is the value of the metric at that timestamp

Example PrometheusMetricsImporter Datasource

Below is an example of one of the built-in ReportDataSource resources that is installed with Metering Operator by default.

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: "pod-request-memory-bytes"
  labels:
    operator-metering: "true"
spec:
  prometheusMetricsImporter:
    query: |
      sum(kube_pod_container_resource_requests_memory_bytes) by (pod, namespace, node)

If the data to be scraped is on a non-default Prometheus instance:

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: "pod-request-memory-bytes"
  labels:
    operator-metering: "true"
spec:
  prometheusMetricsImporter:
    query: |
      sum(kube_pod_container_resource_requests_memory_bytes) by (pod, namespace, node)
    prometheusConfig:
      url: http://custom-prometheus-instance:9090

ReportQuery View Datasource

For ReportDataSources with a spec.reportQueryView present, a Presto view will be created using the rendered output of a specified ReportQuery's spec.query field. This enables abstracting away the details of more complex queries by exposing them as a database table whose content is based on the result of the query the view is based on. It also enables re-use by allowing you to create a view containing the complexities of a query allowing other queries to simply query it as a regular table.

Example ReportQuery View Datasource

This example exposes the pod-memory-request-raw ReportQuery as a view. The schema is based on the spec.columns of the ReportQuery.

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: "pod-memory-request-raw"
  labels:
    operator-metering: "true"
spec:
  reportQueryView:
    queryName: pod-memory-request-raw

If you wanted to specify some inputs to a ReportQuery that accepts inputs, you can set them in the spec.reportQueryView.inputs:

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: "cluster-cpu-capacity-2019"
spec:
  reportQueryView:
    queryName: cluster-cpu-capacity
    inputs:
    - name: ReportingStart
      value: "2019-01-01T00:00:00Z"
    - name: ReportingEnd
      value: "2020-01-01T00:00:00Z"

For more details on how inputs can be specified read the Specifying Inputs section of the ReportQueries documentation.

AWS Billing Datasource

For ReportDataSources with a spec.awsBilling present, see here for an example of what the table schema looks like.

Example AWS Billing Datasource

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: "aws-billing"
  labels:
    operator-metering: "true"
spec:
  awsBilling:
    source:
      bucket: "your-aws-cost-report-bucket"
      prefix: "path/to/report"
      region: "your-buckets-region"

PrestoTable Datasource

For ReportDataSources with a spec.prestoTable present, the reporting-operator will simply verify that a PrestoTable resource exists and it's status.tableName is set. If it does, then the ReportDataSource will simply point at the existing PrestoTable. A PrestoTable ReportDataSource is merely a way to expose an arbitrary table to the rest of the metering resources which expect to interact with a ReportDataSource.

Example PrestoTable Datasource

apiVersion: metering.openshift.io/v1
kind: ReportDataSource
metadata:
  name: example-baremetal-cost
spec:
  prestoTable:
    tableRef:
      name: example-baremetal-cost