Architecture Documents

• Alert History
  • Context
  • Goal
  • User Stories
  • Monitoring and Alerting
  • Deployment
  • Future/Bonus Features
  • Design Choices
    • Alertmanager webhook
    • Loki as storage backend
  • Rejected Design Choices
    • Alertmanager API scraper
    • Dedicated database as storage backend
  • Implementation Details
    • Alertmanager webhook
    • Alertmanager configuration
    • Fluent-bit configuration
    • Expose Loki API
    • Global Health Component Implementation
    • Grafana dashboard
  • Test Plan
• Alerting Functionalities
  • Context
  • Requirements
  • Design Choices
    • Extra Prometheus Rules
    • Edit Existing Prometheus Alert Rules
    • Custom Alertmanager Configuration
    • Retrieve Alert Rules List
    • Silence an Alert
  • Extract Rules Tooling
  • Rejected Design Choices
    • Using amtool vs Alertmanager API
  • Implementation Details
    • Iteration 1
    • Iteration 2
  • Documentation
  • Test Plan
• Metalk8s predefined Alert rules and Alert Grouping
  • Context
  • Vocabulary
  • Requirements
  • Severity Classification
  • Data Model
    • Platform
    • Nodes
    • System Partitions
    • Volumes
    • Platform Services
    • Core
    • Observability
    • Network
    • Access
• Authentication
  • Context
  • Requirements
  • User Stories
    • Pre-provisioned user and password change
    • User Management with local IdP
    • External IDP Integration
    • Authentication check
    • Configuration persistence
    • SSO between Admin UIs
  • Open questions
  • Design Choices
    • Rejected design choices
    • Static password file Vs OpenID Connect
    • X.509 certificates Vs OpenID Connect
    • Keycloak Vs Dex
    • Why not Keycloak?
    • Unexploited choices
  • Implementation Details
    • Iteration 1
    • Iteration 2
    • Iteration 3
    • Iteration 4
    • Iteration 5
  • Documentation
  • Test Plan
• Centralized CLI
  • Context
  • Requirements
    • Cluster Resources Administration
    • Cluster Administration
    • Solution Administration
    • Cluster Service Administration
  • Design Choices
    • metalk8s kubectl plugin
      • Language
      • Input and Output
      • SaltAPI interaction
      • Deployment
      • Rejected design choice
    • metalk8sctl CLI
      • Language
      • Input and Output
      • Salt interaction
      • Deployment
      • Rejected design choice
  • Implementation Details
  • Documentation
  • Test plan
• Continuous Testing
  • Pre-merge
  • Post-merge
    • Nightly
    • Weekly
  • Adaptive test plan
• Cluster and Services Configurations and Persistence
  • Context
  • Requirements
    • User Stories
      • Available Settings
      • Settings Documentation
      • Persistence of Configurations
      • Backup and Restoration
      • Expert-mode Access
  • Design Choices
    • How it works
      • Storage format
      • Non-goals
      • Rejected design choices
      • Consul KV vs ConfigMap
      • Operator (Custom Controller) vs Salt
  • Implementation Details
    • Iteration 1
    • Iteration 2
  • Documentation
  • Test Plan
• Control Plane Ingress
  • Context
  • User Stories
    • MetalK8s and Grafana UIs HA
  • Design Choices
  • Rejected Design Choices
    • Manage Virtual IP by MetalK8s
    • Rely on DNS resolution
    • Use Nodes Control Plane IPs
  • Non-Goals
  • Implementation Details
    • Control Plane Ingress IP
    • MetalLB Configuration
    • Deployment
  • Documentation
  • Test Plan
• Deployment
  • Some notes
    • Configuration
    • Firewall
• Interaction with Kubernetes API from Salt
  • Context
  • Design Choices
  • Rejected Design Choices
    • Migrate to kubectl
    • Contribute to python-kubernetes Lib
    • Direct HTTP Query to Kubernetes API from Python
  • Implementation Details
    • First Iteration
    • Second Iteration (if needed)
  • Documentation
  • Test Plan
• Log Centralization
  • Context
  • Requirements
  • User Stories
  • Deployment & Configuration
  • Monitoring
  • UI
  • Components
    • Collector
    • Distributor (Router)
    • Ingester (Storage)
    • Querier
  • Design Choices
    • Fluent Bit + Loki
      • Fluent Bit
      • Loki
  • Rejected Design Choices
    • Promtail + Loki
    • Fluentd + Loki
    • Fluent Bit + Fluentd + Loki
    • Logging Operator
    • Logstash + Elasticsearch
  • Implementation Details
    • Deployment
      • Fluent Bit
      • Loki
        • Loki Storage
        • Loki HA
    • Configuration
      • Fluent Bit
      • Loki
    • Monitoring
      • Prometheus
      • Grafana
    • Loki Volume Purge
    • Iterations
      • Iteration 1
      • Iteration 2
  • Documentation
  • Test Plan
• Shell-UI
  • Navbar Updater Components
  • Instance Name
• MetalK8s UI adaptation
  • MetalK8s Operations
  • MetalK8s Monitoring
  • MetalK8s Pages
• Monitoring
  • Context
  • Goals
  • Design Design Choices
  • Rejected Choices
    • Federated Prometheus
    • Cortex
  • Implementations Details
  • Documentation
  • Test Plan
• Requirements
  • Deployment
    • Mimick Kubeadm
    • Fully Offline
    • Fully Idempotent
    • Single-Server
    • Scale-Up from Single-Server Deployment
    • Installation == Upgrade
    • Rolling Upgrade
    • Handle CentOS Kernel Memory Accounting
    • At-Rest Encryption
    • Node Labels
    • Vagrant
  • Runtime
    • No Root
    • SELinux
    • Read-Only Containers
    • Environment
    • CRI
    • OIDC
  • Distribution
    • No Random Binaries
    • Tagged Generated Files
    • Container Images
  • Networking
    • Zero-Trust Networking: Transport
    • Zero-Trust Networking: Identity
    • Zero-Trust Networking: Certificate Scope
    • Zero-Trust Networking: Certificate TTL
    • Zero-Trust Networking: Offline Root CAs
    • Zero-Trust Networking: Host Firewall
    • Zero-Trust Networking: No Insecure Ports
    • Zero-Trust Networking: Overlay VPN (Optional)
    • DNS
    • Server Address Changes
    • Multi-Homed Servers
    • Availability of kube-apiserver
    • Provide LoadBalancer Services
    • Network Configuration: MTU
    • Network Configuration: IPIP
    • Network Configuration: BGP
    • IPv6
  • Storage
  • Batteries-Included
    • Metrics and Alerting: Prometheus
    • Node Monitoring: node_exporter
    • Node Monitoring: Platform
    • Node Monitoring: Dashboards
    • Logging
    • Container Registry
    • System Package Repository
    • Tracing Infrastructure (Optional)
    • Backups
• Solutions
  • Context
  • Requirements
  • User Stories
    • Application import
    • Application deployment and lifecycle
    • Multiple instances of an application
    • Application development
    • Application packaging
    • Application validation
  • Design Choices
    • Solutions
      • What’s a Solution
      • Solution Configuration
    • Archive format
    • OCI Images registry
    • Operator Configuration
    • Solution environment
    • Solution management
    • Interaction diagram
  • Rejected design choices
    • CNAB
  • Implementation Details
    • Iteration 1
    • Iteration 2
  • Documentation
  • Test Plan
• Volume Management
  • Abstract
  • Scope
    • Goals
    • Non-Goals
  • Proposal
    • User Stories
      • Volume Creation
      • Volume Monitoring
      • Volume Deletion
    • Component Interactions
  • Implementation Details
    • Volume Status
    • Persistent block device naming
    • Operator Reconciliation Loop
      • Reconciliation Loop (Top Level)
      • Volume Deployment
      • Steady state
      • Volume Finalization
    • Volume Deletion Criteria
  • Documentation
  • Test Plan