The problem beyond one host.
WHAT BREAKS WITH SEVERAL HOSTS
Deciding which host runs what Restarting containers when a host fails Routing traffic to wherever a container is Rolling out updates without downtime Scaling with demand Configuration and secrets across machines
WHAT AN ORCHESTRATOR DOES
Takes a description of the desired state and makes reality match it.
WHAT DESIRED STATE MEANS
How many instances, of which image, with what resources and configuration.
WHAT THE SYSTEM DOES CONTINUOUSLY
Compares reality against it, and corrects differences.
WHY THAT MODEL MATTERS
Failures are handled by the same mechanism as deployments.
WHAT THAT PROVIDES
A container dying is simply a difference to correct.
WHAT ORCHESTRATION COSTS
Substantial complexity Concepts to learn Infrastructure to operate Expertise to diagnose
WHAT TO ASK BEFORE ADOPTING IT
Do we have the problems it solves?
WHAT THOSE PROBLEMS ARE
Several hosts Frequent deployment Variable load A requirement to survive host failure
WHAT IF YOU HAVE NONE OF THEM
You do not need it.
WHAT MOST APPLICATIONS ACTUALLY NEED
One or two servers, running containers, with a proxy.
