outline.txt

Introduction
  - who this is for
  - TOC
  - preparing for hands-on exercises
    - minikube
    - download containers
    - update repo

A brief history of stateful applications on Kubernetes
  - originally just storage
  - replicaset limitations
  - creation of statefulset/petset
  - CRDs & Operators

Trade-offs
  - Management Tradeoffs
    - Elephants vs. Mimmoths
    - custom deployment vs. automated deployment
    - shelling in vs. API
    - shared large db vs. db-per-application
    - dedicated server/VM vs. packing
    - DBA team gating vs. self-service
    - command-and-control vs. templating

  - Peformance trade-offs
    - network latency
    - storage abstraction
    - limits on large databases

  - A comparison of high-availability frameworks
    - Non-containerized
      - regmgr, pgpoolII
    - Paid cloud:
      - RDS, Aurora
    - Advantages of Containers/Kubernetes
    - Containerized solutions:
      - Patroni/Spilo
      - Crunchy Postgres Operator
      - Stolon

  - Real-life use cases: Zalando, bugs.python.org
    - zalando
      - self-service for devs
      - big infrastructure
      - thousands of applications
      - AWS-based cloud
    - bugs.python.org
      - Moved to openshift
      - originally RDS
      - wanted to "own it", save money
      - moved to patroni-based storage

Patroni
  -  “The 80% solution”
    - single-application DBs
    - small database clusters
    - fully automated HA
    - Kubernetes services for LB
    - not federated (but wait)
    - AWS-friendly
  - Patroni vs. Spilo
  - How containerized HA works
    - patroni diagram
  - What’s inside the container
    - interactive tour

Deploying
  - Operator versus manual
    - What's an operator?
    - Why you would use one
      - templating
      - scheduled tasks
    - Manual deployment of Patroni Image
      - interactive exercise
    - Deployment of Operator Image
      - interactive exercise
    OpenShift versus Kubernetes deployments
    - Advantages of Kubernetes
    - Advantages of Openshift
      - Locking limitation

Management
  - Testing HA and failover
    - interactive exercise
  - DevOps-ing database management
    - using operators
    - creating templates for your team
    - workflow for DB deployment
  - Managing database contents
    - yes, use the ORM
  - Administration API
    - Port 8001
    - PatroniCTL
    - API commands
      - Status
      - Failover
      - Scale
      - pg_hba
      - postgres settings (plus some cautions)
  - Operator API
    - Latency
    - ConfigMap
    - Changing things
  - Disaster recovery
    - Backups
      - S3 option per Spilo
      - Auto-restore
      - Other options
        - Building pgBackRest setup for bare metal

Further work
  - Production storage options
    - Gluster
    - Local Storage
  - Public cloud considerations
  - Customizing deployment
  - Contributing to the project
    - Patroni & spilo locations
    - We need docs!
    - Current work:
      - Sync Rep
      - Federation