Snaptastic by Mike Ryan & Thierry Schellenbach (mellowmorning.com)

Snaptastic allows you to automate the snapshotting (backup) and mounting of volumes on your EC2 instances.

It uses tagging of snapshots to figure out which snapshot should be used to populate a volume upon boot.

About the Author

• Thierry Schellenbach, Founder & CTO at Fashiolista
• Mike Ryan, Syadmin at Fashiolista
• Peter van Kampen, Developer at Fashiolista

Install

pip

pip install snaptastic

settings file

Create a settings file /etc/snaptastic/snaptastic_settings.py

AWS_ACCESS_KEY_ID = 'key'
AWS_SECRET_ACCESS_KEY = 'secret'
REGION = 'eu-west-1'

from snaptastic import Snapshotter, EBSVolume


class SolrSnapshotter(Snapshotter):
    name = 'solr'
    
    def get_volumes(self):
        volume = EBSVolume('/dev/sdf', '/mnt/solr', size=20)
        volumes = [volume]
        return volumes

verifying

sudo snaptastic test
sudo snaptastic list-volumes solr

That's it, you're now ready to use snaptastic.

Tutorial

mount

snaptastic mount-snapshots solr

you should now have a mounted volume on /mnt/solr/

lets add a file to it
sudo touch /mnt/solr/helloworld

make snapshots

time to make a snapshot of this important change

snaptastic make-snapshots solr

destroying our work

snaptastic unmount-snapshots solr

check to see if /mnt/solr/ is actually gone

restoring from backups

snaptastic mount-snapshots solr

changing the loglevel for a cronjob

snaptastic mount-snapshots solr --loglevel=WARNING

customizing the tagging

When mounting volumes, snaptastic will search for snapshots with the correct tags. By default it will look for:

• role
• cluster
• environment
• mount point (you're advised not to change this)

To uniquely identify snapshots to application logic. Changing the tags which are used to fit with your application setup is trivial. Simply subclass the snapshotter class and change the get_filter_tags function. Have a look at the example below:

class CustomFilterSnapshotter(Snapshotter):
    name = 'filter_example'
    
    def get_filter_tags(self):
        '''
        The tags which are used for finding the correct snapshot to load from.
        In addition to these tags, mount point is also always added.
        
        Use these to unique identify different parts of your infrastructure
        '''
        tags = {
            'group': self.userdata['group'],
        }
        return tags

settings file

For more examples see examples.py

Snaptastic searches for its setting file at:

• snaptastic_settings on sys.path
• /etc/snaptastic_settings.py
• /etc/snaptastic/snaptastic_settings.py

hooks

Snaptastic defines several hooks to allow custom snapshotting behaviour.

The following hooks are available:

• pre_mounts(self, volumes):
• post_mounts(self, volumes):
• pre_mount(self, vol):
• post_mount(self, vol):
• pre_unmounts(self, volumes):
• post_unmounts(self, volumes):
• pre_unmount(self, vol):
• post_unmount(self, vol):
• pre_snapshots(self, volumes):
• post_snapshots(self, volumes):
• pre_snapshot(self, vol):
• post_snapshot(self, vol):

examples

Basic volume customization

class MySnapshotter(Snapshotter):
    name = 'simple_example'

    def get_volumes(self):
        volumes = [EBSVolume('/dev/sdf1', '/mnt/index', size=200)]
        return volumes

Customizing filter tags

class CustomFilterSnapshotter(Snapshotter):
    name = 'filter_example'
    
    def get_filter_tags(self):
        '''
        The tags which are used for finding the correct snapshot to load from.
        In addition to these tags, mount point is also always added.
        
        Use these to unique identify different parts of your infrastructure
        '''
        tags = {
            'group': self.userdata['group'],
        }
        return tags

Reading volumes from the userdata

class UserdataSnapshotter(Snapshotter):
    '''
    Looks for a list of volumes in the instance's userdata
    [{
       "device": "/dev/sdf1",
       "mount_point": "/var/lib/postgresql/9.1/main",
       "size": 200
    }]
    '''
    name = 'userdata_example'

    def get_volumes(self):
        volume_dicts = self.userdata['volumes']
        volumes = []
        for volume_dict in volume_dicts:
            volume = EBSVolume(**volume_dict)
            volumes.append(volume)
        return volumes

Using the hooks

class PostgreSQLSnapshotter(Snapshotter):
    '''
    Customized mounting hooks for postgres
    '''
    name = 'postgres_example'

    def pre_mounts(self):
        import subprocess
        subprocess.check_output(['/etc/init.d/postgresql', 'stop'])

    def post_mounts(self):
        import subprocess
        # fix permissions on postgresql dirs
        subprocess.check_output(['chmod', '-R', '0700', '/var/lib/postgresql'])
        subprocess.check_output(
            ['chown', '-R', 'postgres:postgres', '/var/lib/postgresql'])

Features

• Graceful failure handling
• Freezes for the absolute minimal required duration
• Batches boto API calls for faster batch volume mounting/snapshotting
• Tested codebase

Porting old systems

You will often need to fake userdata for porting old systems. Doing so is quite easy:

sudo snaptastic make-snapshots solr --userdata='{"role": "solr", "cluster": "solr", "environment": "aws"}'

Cronjobs and logging

Changing the loglevel is really easy. Simply call any of the commands with the --loglevel option. Internally this uses the python logging module. So valid options are: DEBUG,INFO,WARNING,ERROR,CRITICAL For cronjobs I recommend settings it to WARNING

changing the loglevel

snaptastic mount-snapshots solr --loglevel=WARNING

If you need more detailed control over logging you can change the LOGGING_CONFIG dictionary in snaptastic_settings.py used by the python logging module.

The volume object

basic 200GB volume

EBSVolume('/dev/sdf1', '/var/lib/postgresql/9.1/main', 200),

now with IOPS

EBSVolume('/dev/sdf1', '/var/lib/postgresql/9.1/main', 200, iops=1600),

now for EXT4

EBSVolume('/dev/sdf1', '/var/lib/postgresql/9.1/main', 200, file_system=FILESYSTEMS.EXT4),

Todo

• Integrated cleanup scripts
• Auto terminate volumes after instance termination
• Auto detect region for cross region usage

Contributing & Project Workflow

Contributions are more than welcome. Have a look at the Todo to get an idea of what's still missing. Please always add unittests (where possible) for your feature/bug.

• fab validate (checks pep8 and unittests)
• fab publish (if tests are ok, publishes the new version, tag, pypi)
• fab clean (attempt to auto cleanup pep8 mistakes)

Running tests

• python -m unittest snaptastic.tests

Django Jobs

Do you also see the beauty in clean code? Are you experienced with high scalability web apps? Currently we're looking for additional talent over at our Amsterdam office. Feel free to drop me a line at my personal email for more information: thierryschellenbach[at]gmail.com