Dockerize Java: Slim JRE, Fast Startup, No JDK Bloat

Java applications need a JVM, but they do not need a full JDK in production. The openjdk:17 image is over 400 MB — most of which is compiler tooling your app never touches. Meanwhile, memory tuning (-Xmx, -XX:MaxRAMPercentage) is critical inside containers where the JVM can see host memory by default.

Proper containerization means shipping only the JRE modules your app uses, configuring container-aware memory limits, and leveraging layered JARs so dependency layers are cached across builds.

Multi-stage Dockerfile with jlink custom runtime:

• Build stage: FROM eclipse-temurin:21-jdk-alpine AS builder — runs ./mvnw -DskipTests package (or ./gradlew bootJar). For Spring Boot, extracts layers with java -Djarmode=layertools -jar app.jar extract.
• JRE stage: Uses jlink --add-modules $(jdeps --print-module-deps app.jar) to create a custom JRE with only the required modules — typically 60-80 MB instead of 300 MB.
• Runtime stage: FROM alpine:3.19 — copies custom JRE and layered JAR directories (dependencies, spring-boot-loader, application). Sets -XX:MaxRAMPercentage=75.0 and -XX:+UseG1GC.

Compose config includes JVM memory settings via environment variables, a health check on the actuator endpoint, and a stop_grace_period for graceful shutdown.

• Multi-stage Dockerfile with custom JRE — final image 120-180 MB
• Layered JAR extraction for optimal Docker layer caching
• docker-compose.yml with JVM tuning, health checks, and graceful shutdown
• Container-aware memory configuration (MaxRAMPercentage instead of fixed -Xmx)
• CI pipeline with Maven/Gradle dependency cache