FOSS4GNA Day 1 - Workshops
Autoscaling PostgreSQL at Scale (OpenShift)
Slide notes
- Redhat is all in on containers
- distant future: containers for all os services
Docker tips
- clean up after apt-get/yum
- set user running default process to anyting but root
- Docker best practices: https://docs.openshift.com/enterprise/3.0/creating_images/guidelines.html
Kubernetes
PROBLEM: Docker not good at sharing network between contianers
SOLUTION: assemble into a pod to talk over localhost
- FS mounted at pod level
- scheduling affinity configurable with lables
- Makes persistent storage available to entire cluster (persistent volume)
- Service = proxy + load balance
Demo
- Use Kubernetes secrets for sensitive information
Thoughts
- Demo went pretty badly due to user errors/errors in workshop notes, but when working, you can really see how this could work for deploying highly available/scalable apps
- Openshift seems like just a management layer for kubernetes for environment variables, scaling, etc.
- The postgres container we spun up with build in replication was cool, but really just a preconfigured Docker instance with replication and sourcing config for other instances using environment variables.