tobsch-krane 1.0.0

data/lib/krane/kubernetes_resource/pod_disruption_budget.rb

@@ -0,0 +1,23 @@
+# frozen_string_literal: true
+module Krane
+  class PodDisruptionBudget < KubernetesResource
+    TIMEOUT = 10.seconds
+
+    def status
+      exists? ? "Available" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists? && observed_generation == current_generation
+    end
+
+    def deploy_method
+      # Required until https://github.com/kubernetes/kubernetes/issues/45398 changes
+      uses_generate_name? ? :create : :replace_force
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/pod_set_base.rb

@@ -0,0 +1,71 @@
+# frozen_string_literal: true
+require 'krane/kubernetes_resource/pod'
+
+module Krane
+  class PodSetBase < KubernetesResource
+    def failure_message
+      pods.map(&:failure_message).compact.uniq.join("\n")
+    end
+
+    def timeout_message
+      pods.map(&:timeout_message).compact.uniq.join("\n")
+    end
+
+    def fetch_events(kubectl)
+      own_events = super
+      return own_events unless pods.present?
+      most_useful_pod = pods.find(&:deploy_failed?) || pods.find(&:deploy_timed_out?) || pods.first
+      own_events.merge(most_useful_pod.fetch_events(kubectl))
+    end
+
+    def fetch_debug_logs
+      logs = Krane::RemoteLogs.new(
+        logger: @logger,
+        parent_id: id,
+        container_names: container_names,
+        namespace: @namespace,
+        context: @context
+      )
+      logs.sync
+      logs
+    end
+
+    def print_debug_logs?
+      pods.present? # the kubectl command times out if no pods exist
+    end
+
+    private
+
+    def pods
+      raise NotImplementedError, "Subclasses must define a `pods` accessor"
+    end
+
+    def parent_of_pod?(_)
+      raise NotImplementedError, "Subclasses must define a `parent_of_pod?` method"
+    end
+
+    def container_names
+      regular_containers = @definition["spec"]["template"]["spec"]["containers"].map { |c| c["name"] }
+      init_containers = @definition["spec"]["template"]["spec"].fetch("initContainers", {}).map { |c| c["name"] }
+      regular_containers + init_containers
+    end
+
+    def find_pods(cache)
+      all_pods = cache.get_all(Pod.kind, @instance_data["spec"]["selector"]["matchLabels"])
+
+      all_pods.each_with_object([]) do |pod_data, relevant_pods|
+        next unless parent_of_pod?(pod_data)
+        pod = Pod.new(
+          namespace: namespace,
+          context: context,
+          definition: pod_data,
+          logger: @logger,
+          parent: "#{name.capitalize} #{type}",
+          deploy_started_at: @deploy_started_at
+        )
+        pod.sync(cache)
+        relevant_pods << pod
+      end
+    end
+  end
+end

data/lib/krane/kubernetes_resource/pod_template.rb

@@ -0,0 +1,20 @@
+# frozen_string_literal: true
+module Krane
+  class PodTemplate < KubernetesResource
+    def status
+      exists? ? "Available" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/replica_set.rb

@@ -0,0 +1,92 @@
+# frozen_string_literal: true
+require 'krane/kubernetes_resource/pod_set_base'
+
+module Krane
+  class ReplicaSet < PodSetBase
+    TIMEOUT = 5.minutes
+    attr_reader :pods
+
+    def initialize(namespace:, context:, definition:, logger:, statsd_tags: nil,
+      parent: nil, deploy_started_at: nil)
+      @parent = parent
+      @deploy_started_at = deploy_started_at
+      @pods = []
+      super(namespace: namespace, context: context, definition: definition,
+            logger: logger, statsd_tags: statsd_tags)
+    end
+
+    def sync(cache)
+      super
+      @pods = fetch_pods_if_needed(cache) || []
+    end
+
+    def status
+      return super unless rollout_data.present?
+      rollout_data.map { |state_replicas, num| "#{num} #{state_replicas.chop.pluralize(num)}" }.join(", ")
+    end
+
+    def deploy_succeeded?
+      return false if stale_status?
+      desired_replicas == rollout_data["availableReplicas"].to_i &&
+      desired_replicas == rollout_data["readyReplicas"].to_i
+    end
+
+    def deploy_failed?
+      pods.present? &&
+      pods.all?(&:deploy_failed?) &&
+      !stale_status?
+    end
+
+    def desired_replicas
+      return -1 unless exists?
+      @instance_data["spec"]["replicas"].to_i
+    end
+
+    def ready_replicas
+      return -1 unless exists?
+      rollout_data['readyReplicas'].to_i
+    end
+
+    def available_replicas
+      return -1 unless exists?
+      rollout_data["availableReplicas"].to_i
+    end
+
+    private
+
+    def stale_status?
+      observed_generation != current_generation
+    end
+
+    def fetch_pods_if_needed(cache)
+      # If the ReplicaSet doesn't exist, its pods won't either
+      return unless exists?
+      # If the status hasn't been updated yet, we're not going to make a determination anyway
+      return if stale_status?
+      # If we don't want any pods at all, we don't need to look for them
+      return if desired_replicas == 0
+      # We only need to fetch pods so that deploy_failed? can check that they aren't ALL bad.
+      # If we can already tell some pods are ok from the RS data, don't bother fetching them (which can be expensive)
+      # Lower numbers here make us more susceptible to being fooled by replicas without probes briefly appearing ready
+      return if ready_replicas > 1
+
+      find_pods(cache)
+    end
+
+    def rollout_data
+      return { "replicas" => 0 } unless exists?
+      { "replicas" => 0 }.merge(
+        @instance_data["status"].slice("replicas", "availableReplicas", "readyReplicas")
+      )
+    end
+
+    def parent_of_pod?(pod_data)
+      return false unless pod_data.dig("metadata", "ownerReferences")
+      pod_data["metadata"]["ownerReferences"].any? { |ref| ref["uid"] == @instance_data["metadata"]["uid"] }
+    end
+
+    def unmanaged?
+      @parent.blank?
+    end
+  end
+end

data/lib/krane/kubernetes_resource/resource_quota.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+module Krane
+  class ResourceQuota < KubernetesResource
+    TIMEOUT = 30.seconds
+
+    def status
+      exists? ? "In effect" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      @instance_data.dig("spec", "hard") == @instance_data.dig("status", "hard")
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/role.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+module Krane
+  class Role < KubernetesResource
+    TIMEOUT = 30.seconds
+
+    def status
+      exists? ? "Created" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/role_binding.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+module Krane
+  class RoleBinding < KubernetesResource
+    TIMEOUT = 30.seconds
+
+    def status
+      exists? ? "Created" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/secret.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+module Krane
+  class Secret < KubernetesResource
+    TIMEOUT = 30.seconds
+    SENSITIVE_TEMPLATE_CONTENT = true
+    SERVER_DRY_RUNNABLE = true
+
+    def status
+      exists? ? "Available" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/service.rb

@@ -0,0 +1,104 @@
+# frozen_string_literal: true
+require 'krane/kubernetes_resource/pod'
+
+module Krane
+  class Service < KubernetesResource
+    TIMEOUT = 7.minutes
+
+    def sync(cache)
+      super
+      if exists? && selector.present?
+        @related_pods = cache.get_all(Pod.kind, selector)
+        @related_workloads = fetch_related_workloads(cache)
+      else
+        @related_pods = []
+        @related_workloads = []
+      end
+    end
+
+    def status
+      if !exists?
+        "Not found"
+      elsif requires_publishing? && !published?
+        "LoadBalancer IP address is not provisioned yet"
+      elsif !requires_endpoints?
+        "Doesn't require any endpoints"
+      elsif selects_some_pods?
+        "Selects at least 1 pod"
+      else
+        "Selects 0 pods"
+      end
+    end
+
+    def deploy_succeeded?
+      return false unless exists?
+      return published? if requires_publishing?
+      return exists? unless requires_endpoints?
+      # We can't use endpoints if we want the service to be able to fail fast when the pods are down
+      exposes_zero_replica_workload? || selects_some_pods?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      "This service does not seem to select any pods and this is likely invalid. "\
+      "Please confirm the spec.selector is correct and the targeted workload is healthy."
+    end
+
+    private
+
+    def fetch_related_workloads(cache)
+      related_deployments = cache.get_all(Deployment.kind)
+      related_statefulsets = cache.get_all(StatefulSet.kind)
+      (related_deployments + related_statefulsets).select do |workload|
+        selector.all? { |k, v| workload['spec']['template']['metadata']['labels'][k] == v }
+      end
+    end
+
+    def exposes_zero_replica_workload?
+      return false unless related_replica_count
+      related_replica_count == 0
+    end
+
+    def requires_endpoints?
+      # services of type External don't have endpoints
+      return false if external_name_svc?
+
+      # problem counting replicas - by default, assume endpoints are required
+      return true if related_replica_count.blank?
+
+      related_replica_count > 0
+    end
+
+    def selects_some_pods?
+      return false unless selector.present?
+      @related_pods.present?
+    end
+
+    def selector
+      @definition["spec"].fetch("selector", {})
+    end
+
+    def related_replica_count
+      return 0 unless selector.present?
+
+      if @related_workloads.present?
+        @related_workloads.inject(0) { |sum, d| sum + d["spec"]["replicas"].to_i }
+      end
+    end
+
+    def external_name_svc?
+      @definition["spec"]["type"] == "ExternalName"
+    end
+
+    def requires_publishing?
+      @definition["spec"]["type"] == "LoadBalancer"
+    end
+
+    def published?
+      @instance_data.dig('status', 'loadBalancer', 'ingress').present?
+    end
+  end
+end

data/lib/krane/kubernetes_resource/service_account.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+module Krane
+  class ServiceAccount < KubernetesResource
+    TIMEOUT = 30.seconds
+
+    def status
+      exists? ? "Created" : "Not Found"
+    end
+
+    def deploy_succeeded?
+      exists?
+    end
+
+    def deploy_failed?
+      false
+    end
+
+    def timeout_message
+      UNUSUAL_FAILURE_MESSAGE
+    end
+  end
+end

data/lib/krane/kubernetes_resource/stateful_set.rb

@@ -0,0 +1,70 @@
+# frozen_string_literal: true
+require 'krane/kubernetes_resource/pod_set_base'
+module Krane
+  class StatefulSet < PodSetBase
+    TIMEOUT = 10.minutes
+    ONDELETE = 'OnDelete'
+    attr_reader :pods
+
+    def sync(cache)
+      super
+      @pods = exists? ? find_pods(cache) : []
+    end
+
+    def status
+      return super unless @instance_data["status"].present?
+      rollout_data = @instance_data["status"].slice("replicas", "readyReplicas", "currentReplicas")
+      rollout_data.map { |state_replicas, num| "#{num} #{state_replicas.chop.pluralize(num)}" }.join(", ")
+    end
+
+    def deploy_succeeded?
+      if update_strategy == ONDELETE
+        # Gem cannot monitor update since it doesn't occur until delete
+        unless @success_assumption_warning_shown
+          @logger.warn("WARNING: Your StatefulSet's updateStrategy is set to OnDelete, "\
+                       "which means updates will not be applied until its pods are deleted. "\
+                       "Consider switching to rollingUpdate.")
+          @success_assumption_warning_shown = true
+        end
+        true
+      else
+        observed_generation == current_generation &&
+        status_data['currentRevision'] == status_data['updateRevision'] &&
+        desired_replicas == status_data['readyReplicas'].to_i &&
+        desired_replicas == status_data['currentReplicas'].to_i
+      end
+    end
+
+    def deploy_failed?
+      return false if update_strategy == ONDELETE
+      pods.present? && pods.any?(&:deploy_failed?) &&
+      observed_generation == current_generation
+    end
+
+    private
+
+    def update_strategy
+      if exists?
+        @instance_data['spec']['updateStrategy']['type']
+      else
+        'Unknown'
+      end
+    end
+
+    def status_data
+      return { 'readyReplicas' => '-1', 'currentReplicas' => '-2' } unless exists?
+      @instance_data["status"]
+    end
+
+    def desired_replicas
+      return -1 unless exists?
+      @instance_data["spec"]["replicas"].to_i
+    end
+
+    def parent_of_pod?(pod_data)
+      return false unless pod_data.dig("metadata", "ownerReferences")
+      pod_data["metadata"]["ownerReferences"].any? { |ref| ref["uid"] == @instance_data["metadata"]["uid"] } &&
+      @instance_data["status"]["currentRevision"] == pod_data["metadata"]["labels"]["controller-revision-hash"]
+    end
+  end
+end