Strangler Fig in the Real World: Migrating a 12-Year Java Monolith
A patient, year-long modernization of a Java 8 monolith for a regional retail platform, told without the usual triumphalism.
We followed an enterprise platform team for nearly four quarters as they ran a strangler fig migration on a Java 8 monolith that powered their retail catalog. The team did not rewrite the system; they slowly diverted traffic per capability, carefully kept the legacy feature parity, and accepted that some "temporary" adapters would live for two more years. This briefing is a truthful record of that pace, including the four times they paused work entirely, and the architecture diagram that survived the migration.
What this briefing actually contains
- Capability-by-capability migration sequence the team actually followed
- Adapter patterns kept around the legacy authentication path
- A traffic-shifting playbook with their rollback thresholds
- How they protected the existing test surface
- Why they did not adopt a service mesh during the migration
- A skeptical look at the cost-of-delay assumptions they had to defend internally
- The diagram that survived 11 architecture reviews unchanged
What you can take into your team
-
A defensible, capability-first migration order the team can keep using
-
Predictable pause points that did not derail the program
-
A legacy code retirement plan that does not depend on rewrite heroics
₩7,600,000
The fee covers full access to this briefing, the attached retainer notes, and one follow-up question to the responsible editor. Pricing is informational. Engagements are confirmed in writing during the kickoff conversation.
What we are most often asked about this briefing
The strangler fig sequencing is language-neutral, but the adapter chapter is JVM-specific. If you are on a .NET or Ruby monolith, expect to rework the section on classpath isolation.
Reviews — including reservations
It does not romanticize the migration. The chapter on the four pauses is the part I shared with my VP.
Reasonable, sober pacing advice. The diagram pack alone justified the engagement fee.
The bit on why they refused to add a service mesh during the migration is rare and welcome. Most articles assume you will.