1. 4SO
  2. Products
  3. Platform Factory

Platform engineering and private cloud

4SO Platform Factory

Build, import and operate application platforms.

Turns raw servers or existing Kubernetes environments into standardized RKE2, OKD or imported Kubernetes platforms with catalog, fleet, release, policy, disconnected delivery and Day‑2 operations.

  • RKE2
  • OKD
  • Existing Kubernetes
  • Catalog & Fleet
4SO Platform Factory flow: Discover, Model, Compose, Build, Deploy, Operate.
Who it is for
Platform engineering teams and private-cloud operators who run several Kubernetes clusters to one standard.
What it does
Turns raw servers or existing Kubernetes into a standard platform, then runs catalog, fleet, release, upgrade and recovery from one management surface.
Where it runs
A self-contained management plane on RKE2 in your data center, targeting RKE2, OKD or existing Kubernetes. Disconnected delivery runs through the built-in zot registry.
APPROVAL GATE · RECONCILIATIONFrom a deploy request to a platform that stays aligned with desired state
01

The deploy request is recorded and a plan with its impact is built.

01Problem

A private platform is not the sum of disconnected tools.

Kubernetes, registry, identity, policy, monitoring, upgrades and recovery become another operations burden if they are assembled without one product lifecycle.

  1. 01

    Repeatable platforms

    Blueprints and profiles replace environment-specific assembly.

  2. 02

    Keep native behavior

    RKE2, OKD and imported clusters share a product boundary without fake symmetry.

  3. 03

    Day‑2 after install

    Health, drift, node lifecycle and upgrade stay under fleet operations.

  4. 04

    Trace releases

    Desired state and artifacts retain identity and provenance.

02Capabilities

From infrastructure admission to workload delivery.

Factory owns platform lifecycle while target distributions keep their native capabilities.

  1. 01

    Target lifecycle

    Build a managed platform or enroll an existing cluster.

  2. 02

    Profiles and blueprints

    Versioned platform capabilities and templates.

  3. 03

    Health and drift

    Operate many targets through one product model.

  4. 04

    Desired state

    Forgejo records intent; Argo CD reconciles it.

  5. 05

    Artifact plane and offline delivery

    zot keeps canonical OCI identity; oc-mirror v2 + zot provide sealed acquisition.

  6. 06

    Workload delivery

    Application release and environment binding on observed platform capabilities.

  7. 07

    Data protection

    Policy, backup run, restore and drill for managed targets.

  8. 08

    Node and upgrade

    Drain, replace, maintenance and distribution-aware upgrade.

03Deployment model

Management Plane and target distributions remain separate.

The Factory does not migrate itself into the target distribution. RKE2 management stays self-contained while RKE2, OKD and imported Kubernetes are distinct target identities.

TargetEntry pathOwnershipUse
Managed RKE2Factory buildPlatform lifecycle under Product Control PlaneStandard private application platform
Existing KubernetesImport + discoveryExisting runtime + 4SO lifecycle boundaryBring existing clusters under management
Managed OKDDistribution-aware buildOKD owns CVO/MCO/OVN/OLM/SCC behaviorOpenShift-compatible target
Existing OKDImport + capability discoveryNative OKD + 4SO product workflowsManage without duplicate platform stacks
04Lifecycle

Platform lifecycle extends past cluster creation.

A target remains a managed product resource through capability discovery, releases, node change, upgrade and recovery.

  1. 01Discoverprovider / cluster identity
  2. 02Planprofile · blueprint · capability
  3. 03Provision / Importcreate or enroll target
  4. 04Verifyobserved capability state
  5. 05Delivercatalog · release · reconcile
  6. 06Operatehealth · drift · maintenance
  7. 07Upgradedistribution-aware edges
05Profiles

Factory turns distribution, capability and delivery contracts into Platform Profiles.

Each profile combines distribution identity, capability discovery, policy, supply chain and lifecycle. OKD and RKE2 are deliberately not forced into artificial implementation symmetry.

  • RKE2 Platform

    A lightweight, self-contained Kubernetes platform.

    • Managed lifecycle
    • Catalog-driven add-ons
    • Fleet operations
  • OKD Platform

    Uses OKD-native capabilities rather than duplicating them.

    • OVN / OLM / SCC awareness
    • No duplicate platform stacks
    • Managed or imported
  • Imported Kubernetes

    Adopt an existing cluster without rebuilding it.

    • Identity discovery
    • Capability assessment
    • Lifecycle admission
  • Application Platform

    Application delivery over a managed target.

    • Application Release
    • Environment Binding
    • Observed deployment evidence
  • Disconnected

    Acquire, mirror and install without runtime Internet dependency.

    • Controlled acquisition
    • Local registry
    • Verified inventory
  • Fleet

    Operate targets after creation.

    • Node lifecycle
    • Upgrade / maintenance
    • Diagnostics / recovery
Technical reference

Architecture, workflows and operations in detail

For engineers who want to know how the product works underneath: components, boundaries, failure behavior and the operations it runs.

06Architecture

Product authority, desired state, reconciliation and runtime are different layers.

The architecture avoids making Git, Argo or the target cluster an accidental second product source of truth.

  1. AUTHORITY
    PostgreSQL + Platform APIOrg/Project, catalog, fleet, release, durable operations and audit.
  2. DESIRED STATE
    ForgejoVersioned desired configuration and release inputs.
  3. RECONCILIATION
    Argo CDApply desired state and detect drift.
  4. RUNTIME
    RKE2 / OKD / existing KubernetesDistribution-native serving behavior and observed capability state.
LayerComponent / authorityResponsibilityBoundary
Product authorityPostgreSQL + Platform APIOrg/Project, Catalog, Blueprint, Fleet, Release, durable operations and auditProduct and lifecycle authority; not a runtime target
Desired stateForgejoVersioned desired configuration and release inputsDesired state, not runtime truth
ReconciliationArgo CDApplies desired state and detects driftA reconciler; never an independent source of truth
Artifact planezot OCI registryImage/artifact delivery, mirroring and disconnected inventoryArtifact identity and provenance are preserved
Management planeSelf-contained RKE2Control services, identity, PostgreSQL, Git/registry and operator APIsKept separate from the target distribution
TargetsRKE2 · OKD · Existing KubernetesCapability discovery, workload runtime and Day-2 lifecycleDistribution-native capabilities are kept; no duplicate stack is imposed
TECH

Target and delivery contract. Factory does not build a vague abstraction over Kubernetes: product authority, desired state, reconciliation, the artifact plane and the distribution-native runtime keep separate boundaries.

DomainComponent / mechanismBehavior
Managed OKDopenshift-install / oc + Redfish boot mediaBare-metal provisioning and cluster lifecycle follow OKD’s own distribution contract.
Disconnected OKDoc-mirror v2 → zotAcquisition, inventory and mirror are sealed and reviewable before install.
Application runtime traitOptional Dapr sidecar capabilityDapr can serve service invocation, pub/sub, bindings and resiliency, but never takes over product state, workflow or secret authority.
07Workflows and failure

Operations have explicit planning and recovery states.

Risk, approval, retry policy and runtime verification belong to the operation model rather than an informal runbook.

  1. 01PlanDRAFT → PLANNING
  2. 02ApprovalAWAITING_APPROVAL
  3. 03QueueAPPROVED → QUEUED
  4. 04ExecuteRUNNING
  5. 05VerifyVERIFYING
  6. 06SuccessSUCCEEDED
APPROVAL GATE

Deploy through plan and human approval

High-risk actions never reach execution without an independent approval boundary.

REQUEST
↓
PLAN
↓
APPROVAL?
EXECUTE
↙ ↘
BLOCKED
VERIFY
RECONCILIATION LOOP

Desired state ↔ observed runtime

Forgejo intent and Argo reconciliation do not create a second product authority.

FORGEJO
→
ARGO CD
↓
OBSERVE
↑
TARGET
←
DRIFT
DISCONNECTED SUPPLY CHAIN

Acquire → verify → mirror → admit

Artifact identity is sealed before target installation.

ACQUIRE
↓
DIGEST VERIFY
↓
OC-MIRROR V2
ZOT
TARGET ADMIT
  1. Flow 01

    Create Managed Platform

    Raw infrastructure to registered target.

    1. Provider Input
    2. Plan
    3. Preflight
    4. Provision
    5. Distribution Bootstrap
    6. Runtime Verify

    Failure and recoveryPLAN_FAILED remains pre-mutation; unknown side effects require reconciliation.

  2. Flow 02

    Import Existing Cluster

    Enrollment without rebuilding the cluster.

    1. Credential Scope
    2. Identity Discovery
    3. Capability Discovery
    4. Classify
    5. Register
    6. Eligibility

    Failure and recoveryIdentity or capability conflict returns to the operator.

  3. Flow 03

    Application Release

    Release binds desired state to observed runtime.

    1. Artifact Admit
    2. Policy
    3. Desired State
    4. Argo Reconcile
    5. Runtime Observe
    6. Evidence

    Failure and recoveryUnknown outcome is read back before any redispatch.

  4. Flow 04

    Disconnected OKD

    Supply chain is sealed before install.

    1. Acquire
    2. Verify Digest
    3. Inventory
    4. oc-mirror v2
    5. zot Mirror
    6. Target Admit

    Failure and recoveryMissing or substituted artifacts fail closed.

  5. Flow 05

    Upgrade / Node Replace

    Day‑2 remains distribution-aware.

    1. Admission
    2. Drain / Stage
    3. Mutation
    4. Health Read-back
    5. Capability Verify
    6. Return to Fleet

    Failure and recoveryRollback is only used where the underlying workflow supports it.

Additional operational flows

Drift and data protection deserve visible lifecycle paths too.

  1. Flow 06

    Drift Reconciliation

    Resolve desired/observed differences without overwriting distribution-native ownership.

    1. Observe Target
    2. Diff Desired / Observed
    3. Classify Owner
    4. Plan Reconcile
    5. Argo / Native Apply
    6. Capability Read-back
    7. Close Drift

    Failure and recoveryIf the difference belongs to a target-native subsystem, Factory preserves that ownership instead of forcing symmetry.

  2. Flow 07

    Backup / Restore Drill

    Data protection continues through isolated restore and workload verification.

    1. Protection Policy
    2. Capability Check
    3. Create Backup
    4. Verify Artifact
    5. Isolated Restore
    6. Workload Verify
    7. Record Evidence

    Failure and recoveryAn unverified artifact or unverified restore is not accepted as recovery evidence.

ActionAdmission / PreconditionsExecution ownerSuccess criterionFailure / recovery
Create PlatformResolvable profile + provider/target inputsFactory operation executorObserved target matches desired profilePLAN_FAILED / retry / recovery path
Import ClusterIdentity + access + capability discoveryImport workflowRegistered target with observed capabilitiesConflict requires operator resolution
Deploy ApplicationImmutable release + binding + policyForgejo/Argo + bounded executorObserved deployment/service evidenceUnknown result is not blindly retried
Replace NodeExact node identity + safety admissionLeased/fenced operationReplacement ready; old node safely removedLease loss → retry wait or operator
UpgradeCompatibility + health + evidence gateDistribution-aware workflowObserved version/capability healthRollback only where real
Disconnected DeliverySource lock + digest + inventoryacquire/mirror workflowTarget artifact identity matches sealed inventoryMismatch fails closed
FLOW

Failure states: PLAN_FAILED / FAILED / ROLLBACK_FAILED / NEEDS_OPERATOR / CANCELLED. Retry applies only to allowed failure classes; rollback is claimed only where the workflow supports it.

08Operations catalog

Platform engineering becomes a controlled product surface.

The operator works with targets, fleet, releases and lifecycle actions rather than per-cluster tribal procedures.

  • Create platformCreate a managed target from a distribution and profile.
  • Import clusterDiscovery, capability assessment and admission of an existing cluster.
  • Publish blueprintVersion and publish a reusable platform template.
  • Deploy applicationBind an application release to a specific environment.
  • Node lifecycleAdd, drain, remove, replace and remediate.
  • Upgrade platformAdmission, staged execution and observed health.
  • Backup / RestorePolicy, run, drill and restore on the target.
  • Diagnose / ReconcileInspect drift and failure, and return the target to desired state.

Standardize the platform without erasing the distribution.

4SO owns the product lifecycle; the target keeps the native semantics that make it trustworthy.