Switching Between Java Versions Without the Hassle

Working with different projects means dealing with different Java versions. Some of my Hadoop stuff needs Java 8. Metabase plugins want Java 17. I’d start a Spark job and get errors because I forgot to switch JDKs. It happened enough times that I needed a better way.

The Hadoop stack I work with still runs on Java 8. A lot of the libraries expect it. Metabase, on the other hand, uses newer Java features and works better with Java 17. So I keep both installed and switch between them constantly.

At first, I tried the usual route: update-alternatives to flip between versions. It worked, but I’d forget which version was active and waste time debugging compatibility issues that turned out to be just the wrong $JAVA_HOME.

Looking for Something Better

I gave SDKMAN a shot. It’s popular, installs multiple JDKs, and lets you switch with sdk use java x.y.z. Seemed perfect.

But I ran into problems:

• My shell took longer to start because of the initialization.
• It occasionally messed with my environment variables.
• Setting it up on remote servers was more steps than I wanted to remember.

I looked at jEnv and other version managers. They all felt like extra layers. I wanted something that just worked without adding complexity.

The Fix

I wrote a small bash script:

#!/usr/bin/env bash
# switch-java.sh: swap $JAVA_HOME and PATH quickly

case "$1" in
  8)
    export JAVA_HOME="/usr/lib/jvm/java-8-openjdk"
    ;;
  17)
    export JAVA_HOME="/usr/lib/jvm/java-17-openjdk"
    ;;
  *)
    echo "Usage: switch-java.sh [8|17]"
    exit 1
    ;;
esac
export PATH="$JAVA_HOME/bin:$(echo $PATH | sed -e 's#/usr/lib/jvm/java-[0-9]*-openjdk/bin:##g')"
java -version

I made it executable with chmod +x switch-java.sh and put it in my PATH. Now I just run:

switch-java.sh 17

That’s it. I’m on Java 17 for Metabase. Run switch-java.sh 8 and I’m back to Hadoop. The script prints the version so I always know what’s active.