git: reducing repository size (gc and destructive)

| Created | Modified

This article discusses two approaches to reducing the size of a Git repository: non-destructive garbage collection and destructive history rewriting. It provides a script for running git gc on multiple repositories and links to resources for more advanced techniques.

Garbage Collection (non-destructive)

This works especially well when removing a file added in the most recent unpushed commit. Git Garbage Collection automates some of these cleanup jobs.

I ran the following over my source folders:

for gitPath in $(find . -type d -name ".git" -readable -prune -exec realpath {} \; 2>/dev/null); do
    cd $gitPath
    echo ${gitPath}
    # git branch
    sizeBefore=$(du -sh . | cut -f1)
    git fetch -p
    git branch --format "%(refname:short)" | grep -vE "^(develop|master|staging)$"  | xargs git branch -D
    git gc --aggressive --prune=now
    sizeAfter=$(du -sh . |  cut -f1)
    echo "${sizeBefore} -> ${sizeAfter}"
done

Reduce repository size (destructive)

Atlassian has a pretty good article about reducing git repository size. Also take a look at Git Help: removing-sensitive-data-from-a-repository and the GIT BFG.

Trackback