Skip to content

kiratech/krateo-composition-dynamic-controller

 
 

Repository files navigation

Composition Dynamic Controller

The composition-dynamic-controller is an operator that is instantiated by the core-provider to manage the Custom Resources whose Custom Resource Definition is generated by the core-provider.

Summary

Architecture

composition-dynamic-controller architecture

Workflow

composition-dynamic-controller State Diagram

Composition Dynamic Controller (CDC) & Chart Inspector: Secure Helm Lifecycle Management

The Composition Dynamic Controller (CDC) is a specialized Kubernetes operator that orchestrates the end-to-end lifecycle of Krateo compositions. Acting as the reconciliation engine for Composition custom resources, it bridges declarative application definitions with Helm’s packaging system through intelligent automation. The Chart Inspector serves as its "safety advisor," enabling proactive decision-making via dry-run analysis.

Core CDC Workflow (with Chart Inspector Integration)

  1. Reconciliation Trigger

    • Watches for changes to Composition CRs or Helm chart versions.
    • Invokes the Chart Inspector to simulate installations/upgrades before execution.
  2. Dry-Run Analysis Phase (Chart Inspector)

    helm install --dry-run=server <chart> --version <ver>  # Returns:
    • Resource Manifest List: All Kubernetes objects (Deployments, CRDs, etc.) the chart would create along with filename to .
    • Dependency Graph: Order of operations (e.g., CRDs before custom resources).
  3. RBAC Auto-Provisioning (CDC)

    • Dynamically generates least-privilege Roles/ClusterRoles based on the Inspector’s output.
    • Ensures the CDC’s service account has exactly the permissions needed—no more, no less.
  4. Atomic Execution (CDC)

    • Proceeds with helm install/upgrade only after successful dry-run and RBAC setup.

Key Capabilities Enabled by This Collaboration

Feature CDC’s Role Chart Inspector’s Contribution
Version-Sensitive Reconciliation Detects chart version drift; rolls forward/back. Identifies version-specific resource changes during dry-run.
Atomic Upgrades Ensures all-or-nothing upgrades. Pre-flights resource compatibility (e.g., CRD schema changes).
Self-Healing Corrects configuration drift. Provides baseline "desired state" for comparison.
Declarative Enforcement Continuously reconciles actual vs. desired state. Supplies the desired state before cluster changes.
Secure RBAC Generates minimal required permissions. Audits chart manifests for required API operations.

Why This Architecture Matters

  1. Safety Net

    • The Chart Inspector’s dry-run prevents "helm surprises" (e.g., undeclared CRD creations or namespace pollution).
    • Example: Blocks a chart upgrade if the new version requires a ClusterRole the CDC isn’t authorized to manage.
  2. GitOps Compliance

    • The CDC enforces declarative intent by reconciling against the dry-run’s output, not just Helm’s last-applied state.
    • Self-healing kicks in if manual changes violate the composition’s definition.
  3. Multi-Tenancy Ready

    • RBAC is scoped per-composition, isolating teams/projects.
    • The Inspector’s resource listing ensures no cross-tenant leakage (e.g., a composition can’t create resources in forbidden namespaces).

Real-World Example: Handling a Breaking Chart Change

  1. Scenario: A Helm chart v1.2.0 introduces a new CustomResourceDefinition (CRD).
  2. CDC+Inspector Flow:
    • Dry-run detects the new CRD and its required API group permissions.
    • CDC creates a ClusterRole granting create/get/list for the CRD.
    • Upgrade proceeds only after the CRD and RBAC are confirmed active.
  3. Result: Zero downtime; no "helm upgrade failed: CRD missing" errors.

Composition Dynamic Controller Values Injection

The composition-dynamic-controller inject labels and values into the installed resources and in the helm chart release values. This values contains informations about the composition resource associated with the helm release. The values are injected in the following way:

Helm Chart Release Values Installed Resources Labels Description
global.compositionId krateo.io/composition-id The composition resource uid
global.compositionName krateo.io/composition-name The composition resource name
global.compositionNamespace krateo.io/composition-namespace The composition resource namespace
global.compositionInstalledVersion krateo.io/composition-installed-version The version of the composition resource installed. This value changes if the chart version is upgraded
global.compositionApiVersion not injected The api version of the composition resource. This values is deprecated but is mainteined for backward compatibility.
global.compositionGroup krateo.io/composition-group The group of the composition resource.
global.compositionResource krateo.io/composition-resource The plural name of the composition resource.
global.compositionKind krateo.io/composition-kind The kind of the composition resource.
global.krateoNamespace krateo.io/krateo-namespace The namespace where Krateo is installed. This value is used to identify the Krateo resources in the cluster.
global.gracefullyPaused not injected This value is set to true if the annotation krateo.io/gracefully-paused is set on the composition resource. This value is used to pause the reconciliation of the composition resource only after the value is injected in the helm release values with a successful helm upgrade. Read the paragraph below for more details.

About the gracefullyPaused value

The global.gracefullyPaused value provides a way to gracefully pause both the composition resource and all Krateo resources within its Helm chart.

How it works:

  1. Trigger: Set the annotation krateo.io/gracefully-paused on the composition resource
  2. Activation: The pause takes effect only after the next successful Helm upgrade injects this value into the chart
  3. Scope: Pauses both the composition reconciliation AND any Krateo resources in the chart that respect the krateo.io/paused annotation

Use cases:

  • Graceful pause: Temporarily halt all composition-related activity without resource deletion
  • Coordinated pause: Ensure both the composition and its managed resources pause simultaneously
  • Safe maintenance: Pause operations during maintenance windows

Comparison with krateo.io/paused:

Annotation Scope When it takes effect
krateo.io/gracefully-paused Composition + chart resources After next Helm upgrade
krateo.io/paused Composition only Immediately

Example: Use krateo.io/gracefully-paused when you need to pause an entire application stack, or krateo.io/paused for immediate composition-only pausing.

How to include the pause in a resource included in the chart

For Krateo resources that support pausing via the krateo.io/paused annotation:

To include the pause in a resource included in the chart, you can use the krateo.io/paused annotation on the resource. This will ensure that the resource is paused when the composition is paused.

apiVersion: git.krateo.io/v1alpha1
kind: Repo
metadata:
  name: {{ include "fireworks-app.fullname" . }}-repo
  labels:
    {{- include "fireworks-app.labels" . | nindent 4 }}
  annotations:
    krateo.io/paused: "{{ default false (and .Values.global .Values.global.gracefullyPaused) }}"
spec:
...
For non-Krateo resources:

NOTE: Operators implemented without the Krateo runtime may handle "pause" semantics differently (different annotation keys, immediate vs. delayed behavior, or custom fields). Before templating a pause for a non‑Krateo controller, review that operator's API and controller behavior and adapt the Helm template to map global.gracefullyPaused to the operator's expected pause mechanism.

This is an example of how to include the pause in a non-Krateo resource included in the chart. In this case, we use an ArgoCD Application as an example.

apiVersion: argoproj.io/v1alpha1
kind: Application
metadata:
  name: {{ .Release.Name }}
  namespace: {{ .Values.argocd.namespace }}
  labels:
    {{- include "github-scaffolding.labels" . | nindent 4 }}
spec:
  project: {{ .Values.argocd.application.project | default "default" }}
  source:
    repoURL: {{ include "github-scaffolding.toRepoUrl" . }}
    targetRevision: {{ .Values.git.toRepo.branch }}
    path: {{ .Values.argocd.application.source.path }}
  destination:
    server: {{ .Values.argocd.application.destination.server }}
    namespace: {{ .Values.argocd.application.destination.namespace }}

  {{- /* Normalize flags */ -}}
  {{- $hasPaused   := and .Values.global (hasKey .Values.global "gracefullyPaused") -}}
  {{- $paused      := and $hasPaused (eq (toString .Values.global.gracefullyPaused) "true") -}}
  {{- $syncEnabled := default false .Values.argocd.application.syncEnabled -}}

  {{- if $paused }}
  syncPolicy: {}
  {{- else if $syncEnabled }}
  syncPolicy:
    automated:
      prune: {{ default false .Values.argocd.application.syncPolicy.automated.prune }}
      selfHeal: {{ default false .Values.argocd.application.syncPolicy.automated.selfHeal }}
    syncOptions:
      - CreateNamespace=true
  {{- else }}
  syncPolicy: {}
  {{- end }}

The composition-dynamic-controller injects a global.gracefullyPaused boolean into Helm release values after a successful upgrade. When true, chart templates can use this flag to disable automated behavior for non‑Krateo resources (for example emit syncPolicy: {} in an Argo CD Application) and to set krateo.io/paused on Krateo resources, ensuring a coordinated pause across the composition and all included resources.

Configuration

Operator Env Vars

These enviroment varibles can be changed in the Deployment of the composition-dynamic-controller we need to tweak.

Name Description Default Value Notes
COMPOSITION_CONTROLLER_DEBUG dump verbose output false
COMPOSITION_CONTROLLER_WORKERS number of workers 1
COMPOSITION_CONTROLLER_RESYNC_INTERVAL resync interval 3m
COMPOSITION_CONTROLLER_GROUP resource api group populated by core-provider
COMPOSITION_CONTROLLER_VERSION resource api version populated by core-provider
COMPOSITION_CONTROLLER_RESOURCE resource plural name populated by core-provider
COMPOSITION_CONTROLLER_SA_NAME cdc deployment ServiceAccount name populated by core-provider
COMPOSITION_CONTROLLER_SA_NAMESPACE cdc deployment ServiceAccount namespace populated by core-provider
URL_PLURALS NOT USED from version 0.17.1 - URL to krateo pluraliser service http://snowplow.krateo-system.svc.cluster.local:8081/api-info/names Ignored from version 0.17.1
URL_CHART_INSPECTOR url to chart inspector http://chart-inspector.krateo-system.svc.cluster.local:8081/
KRATEO_NAMESPACE namespace where krateo is installed krateo-system
HELM_REGISTRY_CONFIG_PATH default helm config path /tmp
HELM_MAX_HISTORY Max Helm History 3
COMPOSITION_CONTROLLER_MAX_ERROR_RETRY_INTERVAL The maximum interval between retries when an error occurs. This should be less than the half of the poll interval. 0m
COMPOSITION_CONTROLLER_MIN_ERROR_RETRY_INTERVAL The minimum interval between retries when an error occurs. This should be less than max-error-retry-interval. 1m
COMPOSITION_CONTROLLER_METRICS_SERVER_PORT The port where the metrics server will be listening. If not set, the metrics server is disabled.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Go 99.4%
  • Other 0.6%