Fetching contributors…
Cannot retrieve contributors at this time
105 lines (74 sloc) 4.61 KB

Django S3 Cache

Build status Code coverage

This is Amazon Simple Storage Service (S3) cache backend for Django. It is based on the django.core.cache.backends.filebased.FileBasedCache backend and operates in similar fashion. This backend uses django-storages to read/write the data into S3. It uses the s3boto storage backend.

All key/values passed to this backend are stored in a flat directory structure in your S3 bucket. It uses sha1 instead of md5 to create the file name.


Use pip to install from PyPI:

pip install django-s3-cache

Configure the use of this backend:

    'default': {
        'BACKEND': 's3cache.AmazonS3Cache',
        'OPTIONS': {
            'ACCESS_KEY' : 'Your AWS access key',
            'SECRET_KEY' : 'Your AWS secret access key',
            'BUCKET_NAME': 'Your AWS storage bucket name',
            'LOCATION'   : 'directory_prefix',


  • master
    • Switched to Travis-CI
    • Started testing with pylint
    • Internal code refactoring and more testing


Django S3 Cache supports many configuration options. They should be defined as keys of the OPTIONS dictionary in as shown above. If something is not defined explicitly it follows the defaults of s3boto backend from django-storages which in turn reads them from

NOTE-1: some values in may be used globally by boto and other AWS aware Django components since they follow the format AWS_XXXX. It's always best to define your values as cache options explicitly if you don't want to run into problems.

NOTE-2: since version 1.2 Django S3 Cache is compatible with django-storages v1.1.8 which has changed the names of configuration variables. All new variables are expected to be lower case and the AWS keys variables changed names. For exact names see the S3BotoStorage class definition in Django S3 Cache implements backward compatibility with its previous OPTIONS syntax to allow for easier upgrades. Older names are mapped to new ones and all options are lower cased before passing to S3BotoStorage. The example above shows the new syntax.

NOTE-3: before version 1.3 there is a CRITICAL BUG in the handling of the LOCATION option. If used cache objects will be stored under the defined directory, however culling and clearing the cache was not taking this into account. cache.clear() or cache._cull() will delete the entire bucket. This has been fixed in version 1.3!

NOTE-4: in versions 1.2 to 1.3 there is a BUG in the backward compatibility handling of OPTIONS. If you have been using the new style syntax for ACCESS_KEY, SECRET_KEY, BUCKET_NAME it would be overriden and boto will crash due to missing authentication parameters. This has been fixed in version 1.4!

Some notable options are:

  • LOCATION - the directory prefix under which to store cache files. Defaults to empty string, which means the root directory;
  • DEFAULT_ACL == private - default ACL for created objects. Unlike the s3boto storage backend we set this to private;
  • BUCKET_ACL == DEFAULT_ACL - ACL for the bucket if auto created. By default set to private. It's best to use separate bucket for cache files;
  • REDUCED_REDUNDANCY - set to True if you want to save a few cents on storage costs;
  • IS_GZIPPED - set to True to enable Gzip compression. Used together with GZIP_CONTENT_TYPES. See django-storages documentation.

Django S3 implements culling strategy similar to the stock filesystem backend. It will honor the following options:

  • MAX_ENTRIES - the maximum number of entries allowed in the cache before old values are deleted. If 0 culling is disabled. This argument defaults to 300;
  • CULL_FREQUENCY - the fraction of entries that are culled when MAX_ENTRIES is reached. The actual ratio is 1/CULL_FREQUENCY, so set CULL_FREQUENCY to 2 to cull half of the entries when MAX_ENTRIES is reached;


Source code and issue tracker are at