Prototyping Python Microservices in Production

Move fast with stable infra.

Microservices.

• Help move fast, with stable infrastructure.
• Composability.
• Just do one task. Can user service by contract without understanding implementation.
• Reliability.
• Mobility.
• Fault isolation - only if you do it properly.
• Test and deploy at your own pace.
• Ownership. Teams can own individual microservices.
• Safer deployment and testing.
• Scaling. Can move bottlenecks around. Doesn’t actually help.
• Can back fire. Not as easy as it sounds.

Why not?

• Complexity.
• Logging and metrics diverge. Standard logging service. Standard metrics service. Constant checking required.
• Shift in bottlenecks.
• Overheads leads to waste.
• Management to aviod silos.
• Full stack testing is harder.
• Difficult to debug interactively.
• Disaster recovery is daunting.
• Not a silver bullet. Different colour.

Decouple cleanly. Apache Thrift.

• asyncio support coming soon.

Measuring and monitoring. Should be done from the start. Prometheus.

Should not hardcode IP addresses or domain names.

Shadowing. Duplicate requests to instances.