Spring Boot + Python on CloudFoundry (Multi-Buildpack)

A Spring Boot 3.5 / Java 21 application that invokes a Python script via ProcessBuilder. The Python environment (including pandas) is shipped as a conda-pack tarball and unpacked at application startup on CloudFoundry.

What's special about this deployment

Multi-Buildpack

CloudFoundry is configured with two buildpacks in manifest.yaml:

Order	Buildpack	Role
1	python_buildpack	Supply buildpack -- installs a base Python runtime into the droplet
2	java_buildpack_offline	Final buildpack -- provides the JVM, memory calculator and start command for the Spring Boot app

The first buildpack only supplies dependencies; the last one owns the process type and start command.

Conda-Pack for the Python environment

Rather than relying on pip install at staging time, the Python packages (pandas, etc.) are pre-built into a portable conda-pack tarball (python_env.tar.gz). This tarball is created inside a linux/amd64 Docker container so the native binaries match the CloudFoundry stack, even when building on macOS (Apple Silicon or Intel).

At application startup the CF .profile script unpacks the tarball and activates the environment before the JVM starts.

Extracted JAR

The deploy.sh script uses the jar xf command to upack the contents of the JAR file to the deploy directory. This allows the java_buildpack_offline to detect and run the application correctly in a multi-buildpack setup.

Deploy directory layout

After running deploy.sh, the pushed directory looks like this:

deploy/
├── .profile                # CF startup hook -- unpacks conda env
├── requirements.txt        # Empty -- triggers python_buildpack detection
├── runtime.txt             # Python version hint for python_buildpack
├── python_env.tar.gz       # Conda-pack tarball (Python 3.11 + pandas)
├── scripts/
│   └── hello.py            # Python script invoked by the Spring Boot app
└── BOOT-INF/   # Extracted Spring Boot JAR
├── META-INF/
└── org/   

Prerequisites

• Java 21
• Maven (or use the included mvnw wrapper)
• Docker (for building the conda-pack tarball on macOS)
• CF CLI (cf) logged in to a CloudFoundry environment

Build & Deploy

1. Create the conda-pack environment

This runs inside a Docker container so the resulting tarball contains linux-amd64 binaries:

./conda/pack-env.sh

The script produces python_env.tar.gz in the project root.

2. Build the Spring Boot app and assemble the deploy directory

./deploy.sh

This will:

1. Build the Spring Boot fat JAR via Maven
2. Extract it into deploy/
3. Copy the Python script, .profile, requirements.txt, runtime.txt and python_env.tar.gz into deploy/

3. Push to CloudFoundry

cf push

4. Test

curl https://<app-route>/run-python

Expected output:

=== Hello from Python with Pandas! ===
Pandas version: 2.x.x

Language scores:
 language  score
     Java     85
   Python     92
       Go     78
     Rust     88

Average score: 85.8
Top language: Python
=== Done ===