Skip to content

Latest commit

 

History

History
242 lines (174 loc) · 7.37 KB

README.md

File metadata and controls

242 lines (174 loc) · 7.37 KB
Raw

kube-client

Crystal client library for the Kubernetes (1.11+) API

.github/workflows/docs.yml GitHub release Chat on Telegram

Installation

  1. Add the dependency to your shard.yml:

    dependencies:
  kube-client:
    github: spoved/kube-client.cr

  2. Run shards install

Usage

Specify the kubernetes api version of the client to use:

require "kube-client/v1.20"

client = Kube::Client.autoconfig

Or you can specify the kubernetes api version at compile time via the -Dk8s_v{major}.{minor} flag:

require "kube-client"

client = Kube::Client.autoconfig
$ crystal build -Dk8s_v1.20 kube-client.cr

Overview

The top-level Kube::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::Kubernetes::Resource instances, which provide 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 Kube::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 Kube::Error::API subclasses represent different HTTP response codes, such as Kube::Error::NotFound or Kube::Error::Conflict.

Creating a client

Unauthenticated client

client = Kube.client("https://localhost:6443", ssl_verify_peer: false)

The keyword options are Kube::Transport::Options options.

Client from kubeconfig

client = Kube::Client.config(
  Kube::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 = Kube::Client.config(Kube::Config.load_file("~/.kube/config"),
  server: "http://localhost:8001",
)

In-cluster client from pod envs/secrets

client = Kube::Client.in_cluster_config

API Resources

Resources are a sub class of ::K8S::Kubernetes::Resource, which is generated and defined in the k8s.cr sub-shard.

Please note that custom resources are not supported at this time.

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(label_selector: {"role" => "test"}).each do |pod|
  pod = pod.as(K8S::Api::Core::V1::Pod)
  puts "namespace=#{pod.metadata!.namespace} pod: #{pod.metadata!.name} node=#{pod.spec.try &.node_name}"
end

Updating resources

node = client.api("v1").resource("nodes").get("test-node")
node.as(K8S::Api::Core::V1::Node).spec.not_nil!.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(label_selector: {"role" => "test"})

Creating resources

Programmatically defined resources

pod = K8S::Api::Core::V1::Pod.new(
  metadata: {
    name: "pod-name",
    namespace: "default",
    labels: {
      "app" => "kube-client-test",
    },
  },
  spec: {
    containers: [
      {
        name: "test",
        image: "test",
      },
    ],
  }
)

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

pod = client.api("v1").resource("pods").create_resource(pod)

From file(s)

resources = K8S::Kubernetes::Resource.from_file("./test.yaml")

resources = client.create_resources(resources)

Patching resources

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

Watching resources

Watching resources spawns a background fiber that will push K8S::Kubernetes::WatchEvents for the resource onto the returned Channel. A Kube::Error::WatchClosed error will be returned if the watch stream has been closed. A Kube::Error::API error will be returned if an api error is encountered.

resource_client = client.api("v1").resource("pods")
channel = resource_client.watch(resource_version: "4651")

while !channel.closed?
  event = channel.receive
  if event.is_a?(Kube::Error::WatchClosed)
    # Handle if the watch stream has been closed
  elsif event.is_a?(Kube::Error::API)
    # Handle error
  else
    pp event # => K8S::Kubernetes::WatchEvent(K8S::Api::Core::V1::Pod)
  end
end

The returned Kube::Error::WatchClosed error will contain the resource_version of the last event received before the watch stream was closed. This can be used to resume watching from the last known resource version:

resource_client = client.api("v1").resource("pods")
channel = resource_client.watch

while !channel.closed?
  event = channel.receive
  if event.is_a?(Kube::Error::WatchClosed)
    # Restart the watch from the last known resource version
    channel = resource_client.watch(resource_version: event.resource_version)
  else
    pp event # => K8S::Kubernetes::WatchEvent(K8S::Api::Core::V1::Pod)
  end
end

You can also invoke the watch method with a block, which can automatically restart the watch from the last known resource version:

resource_client.watch(auto_resume: true) do |event|
  obj = event.object.as(K8S::Api::Core::V1::Pod)
  Log.info { "#{event.type} #{obj.metadata.try &.name}" }
end

Contributing

  1. Fork it (https://github.com/spoved/kube-client.cr/fork)
  2. Create your feature branch (git checkout -b my-new-feature)
  3. Commit your changes (git commit -am "Add some feature")
  4. Push to the branch (git push origin my-new-feature)
  5. Create a new Pull Request

Contributors