Production engineering begins when software leaves the developer laptop and must survive real traffic, real failures, real operators and real business consequences.

A useful mental model is:

Software correctness
      +
Systems understanding
      +
Operational discipline
      +
Automation
      +
Production feedback
      ↓
Reliable production service

At Meta, Production Engineering is described as a hybrid software-and-systems engineering discipline that champions reliability, scalability, performance and security. Those four words are a strong foundation for this course, but production engineering also requires operability: a service must be observable, diagnosable, deployable, recoverable and supportable by humans.

Production is a different environment

A service that works locally can still fail in production because production introduces:

• concurrency and unpredictable traffic
• partial dependency failures
• real network latency and packet loss
• credentials and authorization boundaries
• capacity limits and quotas
• stale configuration and version skew
• large data sets and migrations
• rolling deployments and mixed versions
• human operators under time pressure

Therefore, the production engineer asks more than “Does it work?”

Can we understand it?
Can we change it safely?
Can it fail partially?
Can it recover?
Can it scale?
Can we prove it is healthy?
Can another engineer operate it at 3 AM?

The four production postures

These postures interact. A cache may improve performance but increase consistency complexity. More replicas may improve availability but increase cost. Aggressive retries may hide small failures but create a retry storm during overload.

Production engineering is therefore trade-off engineering.