Report Issue Edit

Improving performance with caching

We can improve performance by caching dependencies between builds, only re-downloading when necessary. To begin, let’s create a cacheable bundler layer.

Creating the `bundler` layer

To do this, replace the following lines in the build script:

echo "---> Installing gems"
bundle install

with the following:

echo "---> Installing gems"
bundlerlayer="$layersdir/bundler"
mkdir -p "$bundlerlayer"
echo -e 'cache = true\nlaunch = true' > "$bundlerlayer.toml"
bundle install --path "$bundlerlayer" --binstubs "$bundlerlayer/bin"

Your full ruby-buildpack/bin/build script should now look like the following:

#!/usr/bin/env bash
set -eo pipefail

echo "---> Ruby Buildpack"

# 1. GET ARGS
layersdir=$1

# 2. DOWNLOAD RUBY
echo "---> Downloading and extracting Ruby"
rubylayer="$layersdir"/ruby
mkdir -p "$rubylayer"
ruby_url=https://s3-external-1.amazonaws.com/heroku-buildpack-ruby/heroku-18/ruby-2.5.1.tgz
wget -q -O - "$ruby_url" | tar -xzf - -C "$rubylayer"

# 3. MAKE RUBY AVAILABLE DURING LAUNCH
echo -e 'launch = true' > "$rubylayer.toml"

# 4. MAKE RUBY AVAILABLE TO THIS SCRIPT
export PATH="$rubylayer"/bin:$PATH
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+${LD_LIBRARY_PATH}:}"$rubylayer/lib"

# 5. INSTALL BUNDLER
echo "---> Installing bundler"
gem install bundler --no-ri --no-rdoc

# ======= MODIFIED =======
# 6. INSTALL GEMS
echo "---> Installing gems"
bundlerlayer="$layersdir/bundler"
mkdir -p "$bundlerlayer"
echo -e 'cache = true\nlaunch = true' > "$bundlerlayer.toml"
bundle install --path "$bundlerlayer" --binstubs "$bundlerlayer/bin"

# 7. SET DEFAULT START COMMAND
cat > "$layersdir/launch.toml" <<EOL
# our web process
[[processes]]
type = "web"
command = "bundle exec ruby app.rb"

# our worker process
[[processes]]
type = "worker"
command = "bundle exec ruby worker.rb"
EOL

Now when we run:

pack build test-ruby-app --path ./ruby-sample-app --buildpack ./ruby-buildpack

You will see something similar to the following during the EXPORTING phase:

[exporter] Adding layer 'com.examples.buildpacks.ruby:bundler'

Caching dependencies

Now, let’s implement the caching logic. We’ll first need to create a ruby-sample-app/Gemfile.lock file with the contents given below:

Typically you would run bundle install locally to generate this file, but for the sake of simplicity we’ll create ruby-sample-app/Gemfile.lock manually.

GEM
  remote: https://rubygems.org/
  specs:
    mustermann (1.0.3)
    rack (2.0.7)
    rack-protection (2.0.7)
      rack
    sinatra (2.0.7)
      mustermann (~> 1.0)
      rack (~> 2.0)
      rack-protection (= 2.0.7)
      tilt (~> 2.0)
    tilt (2.0.9)

PLATFORMS
  ruby

DEPENDENCIES
  sinatra

BUNDLED WITH
   2.0.2

Replace the gem installation logic from the previous step:

# ...

echo "---> Installing gems"
bundlerlayer="$layersdir/bundler"
mkdir -p "$bundlerlayer"
echo -e 'cache = true\nlaunch = true' > "$bundlerlayer.toml"
bundle install --path "$bundlerlayer" --binstubs "$bundlerlayer/bin"

# ...

with the new logic below that checks to see if any gems have been changed. This simply creates a checksum for the previous Gemfile.lock and compares it to the checksum of the current Gemfile.lock. If they are the same, the gems are reused. If they are not, the new gems are installed.

We’ll now write additional metadata to our bundler.toml of the form cache = true and launch = true. This directs the lifecycle to cache our gems and provide them when launching our application. With cache = true the lifecycle can keep existing gems around so that build times are fast, even with minor Gemfile.lock changes.

# Compares previous Gemfile.lock checksum to the current Gemfile.lock
bundlerlayer="$layersdir/bundler"
local_bundler_checksum=$(sha256sum Gemfile.lock | cut -d ' ' -f 1) 
remote_bundler_checksum=$(cat "$bundlerlayer.toml" | yj -t | jq -r .metadata 2>/dev/null || echo 'not found')

if [[ -f Gemfile.lock && $local_bundler_checksum == $remote_bundler_checksum ]] ; then
    # Determine if no gem dependencies have changed, so it can reuse existing gems without running bundle install
    echo "---> Reusing gems"
    bundle config --local path "$bundlerlayer" >/dev/null
    bundle config --local bin "$bundlerlayer/bin" >/dev/null
else
    # Determine if there has been a gem dependency change and install new gems to the bundler layer; re-using existing and un-changed gems
    echo "---> Installing gems"
    mkdir -p "$bundlerlayer"
    echo -e "cache = true\nlaunch = true\nmetadata = \"$local_bundler_checksum\"" > "$bundlerlayer.toml"
    bundle install --path "$bundlerlayer" --binstubs "$bundlerlayer/bin"
fi

Your full ruby-buildpack/bin/build script will now look like this:

#!/usr/bin/env bash
set -eo pipefail

echo "---> Ruby Buildpack"

# 1. GET ARGS
layersdir=$1

# 2. DOWNLOAD RUBY
echo "---> Downloading and extracting Ruby"
rubylayer="$layersdir"/ruby
mkdir -p "$rubylayer"
ruby_url=https://s3-external-1.amazonaws.com/heroku-buildpack-ruby/heroku-18/ruby-2.5.1.tgz
wget -q -O - "$ruby_url" | tar -xzf - -C "$rubylayer"

# 3. MAKE RUBY AVAILABLE DURING LAUNCH
echo -e 'launch = true' > "$rubylayer.toml"

# 4. MAKE RUBY AVAILABLE TO THIS SCRIPT
export PATH="$rubylayer"/bin:$PATH
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+${LD_LIBRARY_PATH}:}"$rubylayer/lib"

# 5. INSTALL BUNDLER
echo "---> Installing bundler"
gem install bundler --no-ri --no-rdoc

# ======= MODIFIED =======
# 6. INSTALL GEMS
# Compares previous Gemfile.lock checksum to the current Gemfile.lock
bundlerlayer="$layersdir/bundler"
local_bundler_checksum=$(sha256sum Gemfile.lock | cut -d ' ' -f 1) 
remote_bundler_checksum=$(cat "$bundlerlayer.toml" | yj -t | jq -r .metadata 2>/dev/null || echo 'not found')

if [[ -f Gemfile.lock && $local_bundler_checksum == $remote_bundler_checksum ]] ; then
    # Determine if no gem dependencies have changed, so it can reuse existing gems without running bundle install
    echo "---> Reusing gems"
    bundle config --local path "$bundlerlayer" >/dev/null
    bundle config --local bin "$bundlerlayer/bin" >/dev/null
else
    # Determine if there has been a gem dependency change and install new gems to the bundler layer; re-using existing and un-changed gems
    echo "---> Installing gems"
    mkdir -p "$bundlerlayer"
    echo -e "cache = true\nlaunch = true\nmetadata = \"$local_bundler_checksum\"" > "$bundlerlayer.toml"
    bundle install --path "$bundlerlayer" --binstubs "$bundlerlayer/bin"
fi

# 7. SET DEFAULT START COMMAND
cat > "$layersdir/launch.toml" <<EOL
# our web process
[[processes]]
type = "web"
command = "bundle exec ruby app.rb"

# our worker process
[[processes]]
type = "worker"
command = "bundle exec ruby worker.rb"
EOL

Now when you build your app:

pack build test-ruby-app --path ./ruby-sample-app --buildpack ./ruby-buildpack

it will download the gems:

===> BUILDING
[builder] ---> Ruby Buildpack
[builder] ---> Downloading and extracting Ruby
[builder] ---> Installing bundler
...
[builder] ---> Installing gems

If you build the app again:

pack build test-ruby-app --path ./ruby-sample-app --buildpack ./ruby-buildpack

you will see the new caching logic at work during the BUILDING phase:

===> BUILDING
[builder] ---> Ruby Buildpack
[builder] ---> Downloading and extracting Ruby
[builder] ---> Installing bundler
...
[builder] ---> Reusing gems

Next, let’s see how buildpack users may be able to provide configuration to the buildpack.

Next Step

Report Issue Edit

Improving performance with caching

Creating the bundler layer

Caching dependencies

Creating the `bundler` layer