Skip to content

assaydepot/k8s-ruby

 
 

Repository files navigation

K8s::Client provided by k8s-ruby

Readme continues

Ruby client library for the Kubernetes (1.9+) API.

The k8s-ruby project is a fork of kontena/k8s-client.

Unfortunately the company Kontena.io went bankcrupt at the end of the year 2019. They had created many wonderful Ruby projects which we are grateful.

The k8s-ruby library is a community effort to keep k8s-client maintained without any dependencies to the former Kontena.io organization. The library was renamed in order to publish it to Rubygems without conflicting with k8s-client.

Highlights

  • Clean API for dynamic Kubernetes API Groups / Resources
  • Fast API requests using HTTP connection keepalive
  • Fast API discovery and resource listings using pipelined HTTP requests
  • Typed errors with useful debugging information

Installation

Add this line to your application's Gemfile:

gem 'k8s-ruby'

And then execute:

bundle

Or install it yourself as:

gem install k8s-ruby

And then load the code using:

require 'k8s-ruby'

Usage

Overview

The top-level K8s::Client provides access to separate APIClient instances for each Kubernetes API Group (v1, apps/v1, etc.), which in turns provides access to separate ResourceClient instances for each API resource type (nodes, pods, deployments, etc.).

Individual resources are returned as K8s::Resource instances, which are RecursiveOpenStruct instances providing attribute access (resource.metadata.name). The resource instances are returned by methods such as client.api('v1').resource('nodes').get('foo'), and passed as arguments for client.api('v1').resource('nodes').create_resource(res). Resources can also be loaded from disk using K8s::Resource.from_files(path), and passed to the top-level methods such as client.create_resource(res), which lookup the correct API/Resource client from the resource apiVersion and kind.

The different K8s::Error::API subclasses represent different HTTP response codes, such as K8s::Error::NotFound or K8s::Error::Conflict.

Creating a client

Unauthenticated client

client = K8s.client('https://localhost:6443', ssl_verify_peer: false)

The keyword options are Excon options.

Client from kubeconfig

client = K8s::Client.config(
  K8s::Config.load_file(
    File.expand_path '~/.kube/config'
  )
)

Supported kubeconfig options

Not all kubeconfig options are supported, only the following kubeconfig options work:

  • current-context
  • context.cluster
  • context.user
  • cluster.server
  • cluster.insecure_skip_tls_verify
  • cluster.certificate_authority
  • cluster.certificate_authority_data
  • user.client_certificate + user.client_key
  • user.client_certificate_data + user.client_key_data
  • user.token
With overrides
client = K8s::Client.config(K8s::Config.load_file('~/.kube/config'),
  server: 'http://localhost:8001',
)

In-cluster client from pod envs/secrets

client = K8s::Client.in_cluster_config

Logging

Quiet

To supress any warning messages:

K8s::Logging.quiet!
K8s::Transport.quiet!

The K8s::Transport is quiet by default, but other components may log warnings in the future.

Debugging

Log all API requests

K8s::Logging.debug!
K8s::Transport.verbose!
I, [2018-08-09T14:19:50.404739 #1]  INFO -- K8s::Transport: Using config with server=https://167.99.39.233:6443
I, [2018-08-09T14:19:50.629521 #1]  INFO -- K8s::Transport<https://167.99.39.233:6443>: GET /version => HTTP 200: <K8s::API::Version> in 0.224s
I, [2018-08-09T14:19:50.681367 #1]  INFO -- K8s::Transport<https://167.99.39.233:6443>: GET /api/v1 => HTTP 200: <K8s::API::MetaV1::APIResourceList> in 0.046s
I, [2018-08-09T14:19:51.018740 #1]  INFO -- K8s::Transport<https://167.99.39.233:6443>: GET /api/v1/pods => HTTP 200: <K8s::API::MetaV1::List> in 0.316s

Using K8s::Transport.debug! will also log request/response bodies. The EXCON_DEBUG=true env will log all request/response attributes, including headers.

Prefetching API resources

Operations like mapping a resource kind to an API resource URL require knowledge of the API resource lists for the API group. Mapping resources for multiple API groups would require fetching the API resource lists for each API group in turn, leading to additional request latency. This can be optimized using resource prefetching:

client.apis(prefetch_resources: true)

This will fetch the API resource lists for all API groups in a single pipelined request.

Listing resources

client.api('v1').resource('pods', namespace: 'default').list(labelSelector: {'role' => 'test'}).each do |pod|
  puts "namespace=#{pod.metadata.namespace} pod: #{pod.metadata.name} node=#{pod.spec.nodeName}"
end

Updating resources

node = client.api('v1').resource('nodes').get('test-node')

node[:spec][:unschedulable] = true

client.api('v1').resource('nodes').update_resource(node)

Deleting resources

pod = client.api('v1').resource('pods', namespace: 'default').delete('test-pod')
pods = client.api('v1').resource('pods', namespace: 'default').delete_collection(labelSelector: {'role' => 'test'})

Creating resources

Programmatically defined resources

service = K8s::Resource.new({
  apiVersion: 'v1',
  kind: 'Service',
  metadata: {
    namespace: 'default',
    name: 'test',
  },
  spec: {
    type: 'ClusterIP',
    ports: [
      { port: 80 },
    ],
    selector: {'app' => 'test'},
  },
})

logger.info "Create service=#{service.metadata.name} in namespace=#{service.metadata.namespace}"

service = client.api('v1').resource('services').create_resource(service)

From file(s)

resources = K8s::Resource.from_files('./test.yaml')

for resource in resources
  resource = client.create_resource(resource)
end

Patching resources

client.api('apps/v1').resource('deployments', namespace: 'default').merge_patch('test', {
    spec: { replicas: 3 },
})

Watching resources

client.api('v1').resource('pods', namespace: 'default').watch(labelSelector: {'role' => 'test'}) do |watch_event|
  puts "type=#{watch_event.type} pod=#{watch_event.resource.metadata.name}"
end

Contributing

Bug reports and pull requests are welcome on GitHub at k8s-ruby/k8s-ruby.

Releases

No releases published

Packages

No packages published

Languages

  • Ruby 99.9%
  • Other 0.1%