The first line of our local setup guide was a warning that you needed 24 GB of RAM. Running the whole system on a laptop meant starting fourteen services, three databases, a message broker and a gateway, and a new engineer typically spent their first two days getting it to boot before touching a line of business code.
We had arrived at fourteen, one reasonable decision at a time. The system started as three services, split along lines that made sense. After that, every new capability that sounded like a separate concern became a separate service, because that was how the architecture diagrams we admired were drawn, and because splitting felt like the responsible, forward-looking choice. Nobody ever proposed having fourteen. Each service was individually defensible, and nobody was in a position to say that the sum of them no longer was.
The cost was not in the code. It was in everything around the code. Each service had its own pipeline, its own configuration, its own dashboards, its own on-call runbook and its own contract with its neighbors. With five engineers, that meant everyone owned roughly three services outright and had a shallow, partial understanding of the other eleven. A change that crossed a boundary, which was most of the interesting changes, meant touching three repositories, agreeing on a contract version, and deploying in a specific order so that a consumer never met a producer it did not yet understand. A feature that took a day to write took most of a week to ship.
The turning point was a small bug: a rounding rule in a pricing calculation that existed, slightly differently, in two services. Fixing it correctly required changes in both, a coordinated release, and a day spent proving that no consumer in between depended on the old behavior. In a single codebase that would have been one pull request and one test run.
We did not abandon services entirely, which is the part that gets lost in "we went back to a monolith" stories. We asked a narrower question of each of the fourteen: does this piece have a different scaling profile, a different release cadence, or a different owner from everything around it? Two did, one for heavy document processing and one that integrated with an external provider with its own failure behavior. Those stayed separate. The other twelve became modules in a single deployable, with the same boundaries enforced by the build instead of by the network.
Over the following two quarters, the median time from starting a feature to having it in production fell from about nine days to about three. The local environment now starts three processes and fits in 8 GB, and nobody misses the other eleven pipelines.
Microservices were never the mistake. The mistake was treating the number of services as a measure of architectural maturity, when the constraint that actually mattered was how much operational surface five people could hold in their heads. A service boundary is a cost you pay every day, in deployment, observability and coordination, so it needs a reason that pays you back every day too.
More like this?
I write short production war stories like this one, plus a deep-dive architecture series. Subscribe here if you want both in your inbox.
Recommended Resources for Java & Spring Boot Engineers
If you are preparing for Senior/Lead Java interviews or looking to solidify your Spring Boot & Architecture skills, check out these highly-rated resources from the Javarevisited publication (Use promo code friends20 for an exclusive 20% discount automatically applied at checkout):
- Grokking the Java Interview Prepare for core Java, concurrency, JVM internals, and design pattern questions. Get Full Book (Paid) | Download Free Sample Copy
- Grokking the Spring Boot Interview Master Spring Core, Auto-configuration, Spring Data JPA, Security, and Microservices. Get Full Book (Paid) | Download Free Sample Copy
- Grokking the SQL Interview Deep-dive into query optimization, indexing, joins, and complex SQL window functions. Get Full Book (Paid) | Download Free Sample Copy
- The Complete Java + Spring + SQL Interview Bundle Get all three interview guides in a single heavily discounted package. Get the Ultimate Interview Bundle
- Spring Professional Certification Practice Questions (250+ Questions) Validating your skills? Practice with real exam-style questions before taking the Spring Professional certification. Get the Spring Professional Questions | Download Free Sample Copy


Top comments (1)
"One reasonable decision at a time" is exactly how it happens. Nobody ever decides to build 14 services for 5 people; each split looks sensible on its own.
The 24 GB RAM warning in the setup guide is such a telling signal. When onboarding takes two days just to boot the system, the architecture is optimizing for an org chart you don't have yet.
Thanks for writing up the walk-back too. Posts about consolidating are much rarer than posts about splitting, and much more useful for small teams.