Backup Jenkins home periodicallly with git.

jenkins-home-git.sh

#!/bin/bash

# Setup
#
# - Create a new Jenkins Job
# - Mark "None" for Source Control Management
# - Select the "Build Periodically" build trigger
#   - configure to run as frequently as you like
# - Add a new "Execute Shell" build step
#   - Paste the contents of this file as the command
# - Save
#  
# NOTE: before this job will work, you'll need to manually navigate to the $JENKINS_HOME directory 
# and do the initial set up of the git repository.  
# Make sure the appropriate remote is added and the default remote/branch set up.
#  

# Jenkins Configuraitons Directory
cd $JENKINS_HOME

# Add general configurations, job configurations, and user content
git add -- *.xml jobs/*/*.xml userContent/*

# only add user configurations if they exist
if [ -d users ]; then
    user_configs=`ls users/*/config.xml`

    if [ -n "$user_configs" ]; then
        git add $user_configs
    fi
fi

# mark as deleted anything that's been, well, deleted
to_remove=`git status | grep "deleted" | awk '{print $3}'`

if [ -n "$to_remove" ]; then
    git rm --ignore-unmatch $to_remove
fi

git commit -m "Automated Jenkins commit"

git push -q -u origin master

Adding user configurations gave me some errors, and can be simplified:

# only add user configurations if they exist
if [ -d users ]; then
    git add users/*/config.xml
fi

A simpler way of doing deletion resilient to whitespace in file names:
git ls-files -dz | xargs -0r git rm

The initial script works fine for me, but leaves out some more complex (deeper nested) jobs.
So I added
shopt -s globstar
at the beginning and changed
git add -- .xml jobs//.xml userContent/
to
git add -- .xml jobs/**/config.xml userContent/

I used the following to cope with nested jobs folders
find -L jobs -type f -iname config.xml -follow -print0 -exec git add {} \;

This can be simplified:

if [ -d users ]; then
    user_configs=`ls users/*/config.xml`

    if [ -n "$user_configs" ]; then
        git add $user_configs
    fi
fi

to:

ls users/*/config.xml 2>/dev/null | xargs -r git add

And this can be simplified

to_remove=`git status | grep "deleted" | awk '{print $3}'`

if [ -n "$to_remove" ]; then
    git rm --ignore-unmatch $to_remove
fi

to:

git status | awk '/deleted/ { print $3 }' | xargs -r git rm --ignore-unmatch

Final edition:

#!/bin/sh -e

cd "$JENKINS_HOME"

# Add general configurations, secrets, job configurations, nodes, user content, users and plugins info:
ls -1d *.xml secrets/ jobs/*/*.xml nodes/*/*.xml userContent/* users/*/config.xml \
    plugins/*/META-INF/MANIFEST.MF 2>/dev/null | xargs -r -d '\n' git add --

# Track deleted files:
LANG=C git status | awk '$1 == "deleted:" { print $2; }' | xargs -r git rm --ignore-unmatch

LANG=C git status | grep -q '^nothing to commit' || {
    git commit -m "Automated Jenkins commit at $(date '+%Y-%m-%d %H:%M')"
    git push -q -u origin master
}

Published here: https://github.com/ilyaevseev/jenkins2git/

I'm noticing that when you restore jenkins, it's important to make sure the files are all owned by the jenkins user with the jenkins group or you might see the below error in the GUI:

jenkins.model.InvalidBuildsDir: ${ITEM_ROOTDIR}/builds does not exist and probably cannot be created
	at jenkins.model.Jenkins.checkRawBuildsDir(Jenkins.java:3126)
	at jenkins.model.Jenkins.loadConfig(Jenkins.java:3047)
Caused: java.io.IOException
	at jenkins.model.Jenkins.loadConfig(Jenkins.java:3050)
	at jenkins.model.Jenkins.access$1300(Jenkins.java:309)
	at jenkins.model.Jenkins$13.run(Jenkins.java:3145)
	at org.jvnet.hudson.reactor.TaskGraphBuilder$TaskImpl.run(TaskGraphBuilder.java:169)
	at org.jvnet.hudson.reactor.Reactor.runTask(Reactor.java:296)
	at jenkins.model.Jenkins$5.runTask(Jenkins.java:1096)
	at org.jvnet.hudson.reactor.Reactor$2.run(Reactor.java:214)
	at org.jvnet.hudson.reactor.Reactor$Node.run(Reactor.java:117)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
	at java.lang.Thread.run(Thread.java:748)
Caused: org.jvnet.hudson.reactor.ReactorException
	at org.jvnet.hudson.reactor.Reactor.execute(Reactor.java:282)
	at jenkins.InitReactorRunner.run(InitReactorRunner.java:48)
	at jenkins.model.Jenkins.executeReactor(Jenkins.java:1130)
	at jenkins.model.Jenkins.<init>(Jenkins.java:932)
	at hudson.model.Hudson.<init>(Hudson.java:85)
	at hudson.model.Hudson.<init>(Hudson.java:81)
	at hudson.WebAppMain$3.run(WebAppMain.java:233)
Caused: hudson.util.HudsonFailedToLoad
	at hudson.WebAppMain$3.run(WebAppMain.java:250)

transistorgit's comment was mangled by the Markdown renderer, but I think this was the intention:

The initial script works fine for me, but leaves out some more complex (deeper nested) jobs.
So I added
shopt -s globstar
at the beginning and changed
git add -- *.xml jobs/*/*.xml userContent/*
to
git add -- *.xml jobs/**/config.xml userContent/*

This script has one issue though, folders and build history are not saved.