Skip to content

Latest commit

 

History

History
108 lines (78 loc) · 3.99 KB

datasource.md

File metadata and controls

108 lines (78 loc) · 3.99 KB
Raw

Datasources

A Projection Mapping Datasource is something that defines how to turn a source file (configuration file) into a projected output file in a given format. Datasources can be "projected" into a set of data that will be used to build a ConfigMap. Each datasource can result in 1 or more projected items in the ConfigMap.

NOTE: a Datasource is part of the larger schema for projection manifests. A datasource only makes sense in a projection mapping file.

Schema

A datasource has the following schema:

source: "some/file.json"
output_file: "myconfig.json"
source_format: file|glob|yaml|json
# note: only one of extract, field_extractions may be used
extract: "$.json.path[2].notation.scalar"
field_extractions:
- hostname: "$.data.hostname"
- rack_position: "$.data.rack_position"
output_format: raw|json|yaml

Source

A source is a relative path to the --config-repo argument. The file here is loaded at projection time, and used to extract fields with either extract, field_extractions, or injected wholesale into the ConfigMap when output_format: raw.

It is useful to use structured (yaml or json) sources, to enable surgical field extraction. This allows downstream consumers of the produced ConfigMap to take a least-surface-area approach to configuration. Alternatively, files can be injected wholesale with output_format: raw.

Source Types

Supported sources are:

  • file: This is a raw, unstructured file. You cannot extract fields from this source. This is default if you omit both extract and field_extractions.
  • json: Enables structured field extraction. This is inferred if source ends in .json.
  • yaml: Enables structured field extraction. This is inferred if source ends in .yaml.
  • glob: If source contains *, this is assumed. No structured field extraction capability, but this allows you to project multiple files into your ConfigMap

Output File

This is what the ConfigMap's data is named. This will be used as the projected filename for a ConfigMap when used as a Volume in Kubernetes.

Its useful to make this field's name reflect the type of data contained. I.e. if you are doing field_extractions, you should name this so it reflects the output_format

Extract

This uses jsonpath notation to pull out a scalar value from the source, and make its value the data of this DataSource when projected. See https://github.com/oliveagle/jsonpath for docs on the syntax.

Example source:

foo:
  bar:
  - hello
  - world
nodes:
- hostname: foo-12345.domain.tld
  ip: 1.2.3.4
- hostname: bar-56849.domain.tld
  ip: 2.3.4.5

Example field extraction for foo-12345: $.nodes[0].hostname

Field Extractions

This is the same as above, but with the ability to pull out multiple fields and create a new structured output (yaml or json) with only some keys. Keeping the above source, we can create a new YAML output containing only fields we care about

source: above/file.json
output_format: yaml
output_file: something.yaml
field_extractions:
  # lets pull out "world" as scope
  scope: $.foo.bar[1]
  # this can extract and project whole structured subsets! note that we dont pull out both fields, but the whole map here
  node: $.nodes[0]

Resulting configmap:

kind: ConfigMap
apiVersion: v1
metadata
  name: something
  namespace: some-namespace
  labels:
    tumblr.com/config-version: "1539778254"
    tumblr.com/managed-configmap: "true"
  data:
    something.yaml: |
      scope: "world"
      node:
        hostname: foo-12345.domain.tld
        ip: 1.2.3.4

Output File

This determines the key in data: in the ConfigMap. This will be the "filename" your configmap's data elements project onto the filesystem as. This should match your output_format suffix.

Output Format

  • raw: just the unadulterated file
  • json: inferred when output_file is *.json. Will format structured data in JSON.
  • yaml: inferred when output_file is *.yaml. Will format structured data in YAML.