TechPrep
Worked promptsReadPracticeCheck

Worked prompt: a Principal-level migration strategy

Move a revenue-critical monolith to a target architecture across many teams without a freeze: the design is the plan.

This prompt was written for this site to practise the method. It is not a reported interview question from any company.

Prompt. A .NET monolith serves 80% of revenue. Twelve teams commit to it, releases are monthly and risky. Leadership wants "microservices". You are the Principal engineer. What do you propose?

1. Reframe the goal

"Microservices" is a means. The outcomes are probably deployment frequency, lead time, change failure rate and team autonomy (the DORA measures). Agree the target metrics first; they decide how far to go.

2. Understand today

Map the domain (event storming with the teams), the coupling hot spots from change history (which files change together), the data ownership, and the deployment pipeline. Often the fastest win is the pipeline, not the architecture.

3. Options

Modular monolith first: enforce module boundaries and data ownership inside one deployable; cheap and reversible.

Strangler fig: route specific capabilities to new services behind the same front door, one at a time.

Big-bang rewrite: name it to reject it, with reasons (risk, frozen features, second-system effect).

4. The proposal

Phase 0: CI/CD and trunk-based development with feature flags, so the monolith ships weekly. Phase 1: modularise along the domain map, each module owned by one team, no shared tables. Phase 2: extract the modules with the strongest case (different scaling, different release cadence, compliance boundary) behind the strangler, starting with a low-risk one to build the platform (observability, service templates, contract tests).

5. Risks and governance

Distributed-monolith risk: extracted services that must deploy together. Guard with contract tests and an architecture decision record per extraction.

Data migration is the riskiest part: dual-writes or change-data-capture with reconciliation, never a weekend cutover of the revenue path.

Org design: team boundaries should match service boundaries (Conway's law); propose the team changes explicitly.

6. How you would drive it

A one-page strategy, a quarterly roadmap with measurable exits, an architecture review forum that teaches rather than gates, and a platform team that makes the right path the easy path.

Practice

Written for this site to drill the lesson. Not reported interview questions.

  1. The CFO asks for the cost and the payback period of the programme. Structure the answer.
  2. One team refuses to stop writing to another team's tables. How do you resolve it without escalating first?
  3. Write the architecture decision record for the first extraction.

Check yourself

Why start with a modular monolith?

It establishes boundaries and ownership cheaply and reversibly before paying the distributed-systems cost.

What is the riskiest part of most extractions?

Moving data ownership: dual-writes or CDC with reconciliation.

Sources

Saved in this browser.