Capstone: Scheduler, Everything

The finale of the Distributed Scheduler series, folding every earlier board into one program, alongside the project lesson Build Your Own Distributed Scheduler - From Scratch.

Before you start. This lesson assumes you built the small scheduler across the seven lessons before it, so here is what it can already do, one line each:

• It writes a job down as a rule with a time zone, in a table that survives a restart, instead of a variable a crash erases.
• It works out the exact next moment a job is due, freshly, in the zone the rule names.
• It keeps every job’s next-due moment in one ordered pile, so it can always name the soonest one without checking every job in turn.
• It sleeps until that exact moment, wakes up, and finds the job due.
• It runs several copies at once for safety, but lets only one of them act on any job at a time.
• It hands the actual work to separate helper programs, called workers, instead of doing the work itself.
• It survives a worker doing the same job twice by accident, so the customer only ever sees it once.