Skip to content

Latest commit

 

History

History
477 lines (360 loc) · 10.8 KB

README.md

File metadata and controls

477 lines (360 loc) · 10.8 KB
Raw

Kubernetes / openshift templating and deployment engine

Simple python templating and deployment engine for openshift/k8 yml files. Detect changes made and only applies the required objects.

Use case

This tool was born in the need to a simple templating system which can track changes (similar to terraform) but with k8s object awareness. In addition, it was important for us to have proper native yml support, including object merging - something that was missing in other templating engines.

Installation

  • Python 3.9 or later
  • kubectl or oc binary in path
pip install octoploy

Usage

In octoploy you define apps, each app can contain multiple k8s yml files. Additionally, there is a root configuration which describes the k8s namespace, context, defines global variables and references libraries.

All yml files will be pre-processed before they will be imported. This includes replacing any known ${KEY} variables with their associate values and merging referenced templates.

graph  TB
    root[Root folder<br>_root.yml]
    root --> app1[App 1<br>_index.yml<br>k8s yml files]
    root --> app2[App 2<br>_index.yml<br>k8s yml files]
    
    lib[Library folder<br>_root.yml]
    lib --> common[Common<br>_index.yml<br>k8s yml files]
    
    root --> lib
Loading

Deploy / Plan

Deploys everything

octoploy deploy-all

Deploys all object of the app with the give name

octoploy deploy nginx

The same commands are available for plan - which will list changes to be applied.

octoploy plan / plan-all

This command executes the on-config-change trigger

octoploy reload prometheus

Folder structure

octoploy
|- _root.yml <- Root config, describes the k8s namespace and variables
|- my-app <- Any folder with an _index.yml file is considered an app
    |- _index.yml <- App config
    |- deployment.yml <- K8s yml file(s)
    |- secrets.yml
    
Using this structure you could run `octoploy deploy my-app` to deploy
all yml files inside the my-app folder.

Root config

Here is a sample _root.yml file

# Name of the k8s namespace / openshift project
namespace: 'my-project'

# K8s context which should be used for deploying
context: 'my-k8s-cluster-config'

#####################
# OPTIONAL SECTION
#####################

# Name of the configmap which should hold the octoploy state
stateName: 'octoploy-state'

# Global variables
vars:
  DOMAIN: "dev-core.org"

# Inherit libraries
libraries:
  - common

App config

An app is represented by a folder containing an _index.yml file. The following shows all available parameters.

# Name of the deployment, available as APP_NAME variable
name: 'my-app'

#####################
# OPTIONAL SECTION
#####################

# The type defines how the app will be used.
# Can be "app" (default) or "template"
type: 'app'

# Template which should be applied, none by default
applyTemplates: [ ]

# Templates which should be applied AFTER processing the other templates and base yml files
postApplyTemplates: [ ]

includes:
  # List of additional k8s files that should be included
  # and are located outside of this app folder.
  # All paths are relative to this config file.
  k8s: [ ]

# Action which should be executed if a configmap has been changed
on-config-change:
  # Available options: 
  # deploy (triggers a restart of the deployment)
  - deploy

  # exec (Executes a command inside the running container)
  - exec:
      command: /bin/sh
      args:
        - "-c"
        - "kill -HUP $(ps a | grep prometheus | grep -v grep | awk '{print $1}')"

# Additional variables which are used for replacements
vars:
  NSQ_NAME: 'nsq-core'

# Required parameters which must be passed to this app.
# This can be done via the CLI or if this app is a template the referring app can define it in "vars" 
params:
  - PASSWORD

# File based configmaps
# More modes available, see below
configmaps:
  - name: nginx-config
    files:
      - file: "nginx.conf"

Configmaps

In addition to the regular configmaps you can also define configmaps with a file source. This is done in the _index.yml file:

configmaps:
  - name: nginx-config
    files:
      - file: "nginx.conf"

This will create a new configmap from the file nginx.conf with the name nginx-config. Any changes made to the file will be automatically deployed.

Variables

You can refer to variables in yml files by using ${VAR-NAME}. Variables can also be loaded from files (see loaders below).

# _index.yml
vars:
  # Regular key/value assignment
  key: value

It is also possible to decorate objects using variables and merge them into other objects:

# _index.yml
vars:
  someMerging:
    hello: world
    replicas: 2
# deployment.yml
spec:
  replicas: 1
  _merge: ${someMerging}

Results in:

spec:
  replicas: 2
  hello: world

Global variables

The following variables are available anywhere inside the yml files by default

Key Value
APP_NAME Name of the app in the _index.yml
NAMESPACE Name of the namespace in _root.yml

Value Loaders

You can load values from various sources using value loaders

vars:
  # This will load the public/private and intermediate certs
  # from a pem file and store it in *_KEY, *_PUBLIC, *_CACERT
  # where * is the key of the value.
  # In this example: CERT_KEY, CERT_PUBLIC, CERT_CACERT
  CERT:
    loader: pem
    file: my-cert.pem

  # Load the content of a file into "MY_FILE"
  MY_FILE:
    loader: file
    file: someFile.bin
    # Optional: Defines in which encoding the file content should be read
    # utf-8 by default, only applicable if "conversion" is not set.
    encoding: utf-8

    # Optional: Conversion can be used to convert a binary file 
    # into a string representation, in this base base64 
    conversion: base64

  # Load the content of environment variables into ENV_*
  ENV:
    loader: env

App Templates

You can use templates to reuse and generate yml files. This might be useful when you want to attach sidecar containers to multiple apps. To do so you create a new app with the type field set to template. Other apps can now refer to this template via the applyTemplates or postApplyTemplates field. Templates can refer to other templates (recursively). Any vars defined are passed to the next template.

classDiagram
    App <|-- Root: App receives global variables
    Template <|-- App: Template receives variables
    
    Template --|>App: Injects yml files into app
    class Root{
      +variables
    }
    class App{
      +parameters
      +k8s yml files
    }
    class Template{
      +parameters
      +k8s yml files
    }
Loading

Example:

|- some-template
    |- _index.yml
    |- dc.yml

|- my-app
    |- _index.yml <-- Refers to "some-template"
    |- others.yml

Will result in

|- my-app
    |- _index.yml
    |- others.yml
    |- dc.yml

Object merging

The template engine support content aware object merging. This allows you to decorate existing templates or enhance apps with features. A common example would be a template which adds a monitoring sidecar container. In the examples you can find the nsq-template which defines a sidecar container.

Here is a minimal example:

# examples/nsq-template/dc.yml
kind: Deployment
apiVersion: apps/v1
spec:
  template:
    spec:
      containers:
        - name: "nsqd"
        image: "nsqio/nsq"
# examples/my-app/dc.yml
kind: Deployment
apiVersion: apps/v1
metadata:
  name: "${APP_NAME}"

spec:
  replicas: 1
  selector:
    name: "${APP_NAME}"
  strategy:
    type: RollingUpdate

  template:
    metadata:
      labels:
        name: "${APP_NAME}"

    spec:
      containers:
        - name: "${APP_NAME}"
          image: "docker-registry.default.svc:5000/oc-project/my-app:prod"

If we now apply the nsq-template to our app using postApplyTemplates: [nsq-template] the DeploymentConfig object gets automatically merged:

# Merged result after applying template
kind: Deployment
apiVersion: apps/v1
metadata:
  name: "${APP_NAME}"

spec:
  replicas: 1
  selector:
    name: "${APP_NAME}"
  strategy:
    type: RollingUpdate

  template:
    metadata:
      labels:
        name: "${APP_NAME}"

    spec:
      containers:
        - name: "${APP_NAME}"
          image: "docker-registry.default.svc:5000/oc-project/my-app:prod"
        # This is the part of the template
        - name: "nsqd"
          image: "nsqio/nsq"

Loops

Loops allow you to apply the same template with different parameters. This is useful when deploying microservices which all have the same openshift config.

# _index.yml
applyTemplates: [ api-template ]
forEach:
  # NAME is required for each instance that should be created
  - NAME: entity-compare-api
    # You can define other vars as well
    PORT: 8080

  - NAME: favorite-api
    PORT: 8081

Library

It's possible to define whole project as a library. This allows all apps and templates to be reused by another project. An example would be the same setup for multiple systems which are separated by projects (e.g. dev/test/prod).

# testLib/_root.yml
type: library

# Required parameters
params:
  - domain
  - dockerDomain
  - imageStreamTag
# prod/_root.yml
namespace: 'prod-project'
libraries: 
  - testLib

# Required parameters
vars:
  domain: my-prod.dev-core.org
  dockerDomain: prod-docker.com
  imageStreamTag: prod

When you now deploy the prod project it will inherit all apps inside testLib.

In case you only want to inherit some of the apps from the library, you can disable certain apps using:

# prod/_root.yml
namespace: 'prod-project'
libraries: 
  - testLib

apps: # We don't want to use "app1" of the inherited library
    app1:
      enabled: false

Secrets

You can encrypt your k8s secrets using

export OCTOPLOY_KEY="my password"
octoploy encrypt secrets.yml

The file will be updated in place.

For deploying encrypted secrets, you'll need to set the environment variable OCTOPLOY_KEY with your key used to encrypt the data.

State tracking

Octoploy currently uses a ConfigMap called octoploy-state to keep track of the object states.

The ConfigMap contains all managed objects and their md5 sum. If this hash has changed the whole object will be applied. If the object does already exist, but is not listed in the state it will simply be added to the state.

You can modify the name of the configmap by setting the stateName variable in the _root.yml file.

Examples

All examples can be found in the examples folder.

Contribute

The code should be mostly commented. If you found a bug or want to improve something feel free to open an issue and discuss your ideas.