Log Archiving in Rails


Archiving your logs to Amazon S3 is a simple way to backup your log files. They'll be on S3 if you ever need them and you'll free up space on you servers. You could use a logging service, but if you're like me you won't be using the fancy interface. Ruby has a built in solution to log rotation. The only other things you need is the rake task to archive the files to S3 and a Systemd Timer (or use the crontab).

The Logger

The Rails ActiveSupport::Logger is a wrapper over Ruby's built in Logger. The Ruby logger has an option for rotating the logs based on size or time.

To create a new logger that rotates based on time:

logger = Logger.new('production.log', 'daily') # or 'weekly', 'monthly'

The time based logger rotation didn't seem to work to well for me though, so I ended up using a sized based rotation and haven't had any issues with it.

To create a new logger based on time:

logger = Logger.new('production.log', 10, 104857600)   # Roteate logs when it is 100MB in size and keep 10
logger = Logger.new('production.log', 1000, 104857600) # Roteate logs when it is 100MB in size and keep 1000)

To add support for Rails we just modify the environment file (config/environments/production.rb) and pass the same arguments to ActiveSupport::Logger which will then be passed on to the standard Ruby Logger.

Rails.application.configure do

    # Your production configuration.

    # Roteate logs when it is 100MB in size and keep 100.
    config.logger = ActiveSupport::Logger.new(config.paths['log'].first, 100, 104857600)

end

The configs.paths['log'].first will grab the file used for logging based off the current environment (eg. log/production.log).

The Rake Task

The rake task is what will clean up any of the rotated log files and archive them to S3. It looks for all rotated log files and then uses s3cmd to upload them to an Amazon S3 bucket. Once completed, the task removes the rotated logs. The task uses the AWS credentials in a configuration file (config/application.yml in this case).

The full rake task with comments (lib/tasks/log.rake):

namespace :log do

  desc "Archive Roated logs to S3"
  task :archive do

    config = Rails.application.config_for(:application)['aws']

    # Find all of the log files that have been rotated
    files = Dir.entries("#{Rails.root}/log").select { |x| x =~ /\A.+\.log\.\d+\z/ }

    # Compress the log files
    files.each { |file| `/usr/bin/gzip #{Rails.root}/log/#{file}` }
    gziped_files = Dir.entries("#{Rails.root}/log").select { |x| x =~ /\A.+\.log\.\d+\.gz\z/ }

    # Write the configuration for s3cmd
    f = File.new("#{Rails.root}/tmp/.s3cfg", 'w')
    f << (<<-S3CONFIG).strip_heredoc
        [default]
        access_key = #{config['access_key_id']}
        bucket_location = US
        encoding = UTF-8
        guess_mime_type = True
        host_base = s3.amazonaws.com
        host_bucket = %(bucket)s.s3.amazonaws.com
        secret_key = #{config['secret_access_key']}
        use_https = True
    S3CONFIG
    f.close

    # Store the compressed files to S3
    hostname = `/usr/bin/hostname`.strip
    app_name = Rails.application.class.name.split('::')[0..-2].join('::')
    log_bucket = 'AWS_BUCKET' # Your AWS bucket for the logs
    gziped_files.each do |file|
      time = Time.now.utc.iso8601
      destination = file.gsub(/\A(.+)\.log\.(\d+)\.gz\z/, "s3://#{AWS_BUCKET}/#{app_name}/\\1/#{time}-#{hostname}.\\2.log.gz")
      `s3cmd --config '#{Rails.root}/tmp/.s3cfg' put '#{Rails.root}/log/#{file}' '#{destination}'`
      `rm '#{Rails.root}/log/#{file}'` if $? == 0
    end
  end

end

And example configuration file:

production:
  aws:
    access_key_id:  'ACCESS_KEY'
    secret_access_key:  'SECRET_KEY'
    asset_bucket:  'BUCKET'
    asset_bucket_host_alias: 'BUCKET_ALIAS'

The Systemd Timer

The last piece is to setup the periodic task to archive all rotated logs. The crontab is an alternative if you aren't familiar with Systemd. Most systems are now using systemd, and if you are using systemd units to controll your app, it's the way to go.

There are two files to create to setup the timer, the service file and the timer file. Place the following into the /usr/lib/systemd/system/APPNAME-log-archiver.timer.

[Unit]
Description=APPNAME Log Archiver

[Timer]
# How often you want the logs to be archived
# See: http://www.freedesktop.org/software/systemd/man/systemd.time.html
OnCalendar=daily

[Install]
# Use WantedBy to add as a dependency to the application
# WantedBy=app.target

Then create the related service for the timer /usr/lib/systemd/system/APPNAME-log-archiver.service.

[Unit]
Description=APPNAME Log Archiver
After=network.target

[Service]
# I tend to have a conf file for my Rails App Systemd Servcie for specifing
# the environment to run in. Depends on how your setup is though.
# .include /usr/lib/systemd/system/core.target.conf
User=APPUSER
Group=APPUSER
# The working directory for your Rails app, my apps are all served
# from the /srv/APPNAME/current directory (Capistrano setup)
WorkingDirectory=/srv/APPNAME/current
ExecStart=/usr/bin/bundle exec rake log:archive

Enable

Once you deploy the app you can then enable the systemd timer on each server.

systemctl enable APPNAME-log-archvier.timer
systemctl start APPNAME-log-archvier.timer