This tutorial focuses on the boto interface to the Simple Storage Service from Amazon Web Services. This tutorial assumes that you have already downloaded and installed boto.
The first step in accessing S3 is to create a connection to the service. There are two ways to do this in boto. The first is:
>>> from boto.s3.connection import S3Connection >>> conn = S3Connection('<aws access key>', '<aws secret key>')
At this point the variable conn will point to an S3Connection object. In this example, the AWS access key and AWS secret key are passed in to the method explicitely. Alternatively, you can set the environment variables:
and then call the constructor without any arguments, like this:
>>> conn = S3Connection()
There is also a shortcut function in the boto package, called connect_s3 that may provide a slightly easier means of creating a connection:
>>> import boto >>> conn = boto.connect_s3()
In either case, conn will point to an S3Connection object which we will use throughout the remainder of this tutorial.
Once you have a connection established with S3, you will probably want to create a bucket. A bucket is a container used to store key/value pairs in S3. A bucket can hold an unlimited amount of data so you could potentially have just one bucket in S3 for all of your information. Or, you could create separate buckets for different types of data. You can figure all of that out later, first let’s just create a bucket. That can be accomplished like this:
>>> bucket = conn.create_bucket('mybucket') Traceback (most recent call last): File "<stdin>", line 1, in ? File "boto/connection.py", line 285, in create_bucket raise S3CreateError(response.status, response.reason) boto.exception.S3CreateError: S3Error: Conflict
Whoa. What happened there? Well, the thing you have to know about buckets is that they are kind of like domain names. It’s one flat name space that everyone who uses S3 shares. So, someone has already create a bucket called “mybucket” in S3 and that means no one else can grab that bucket name. So, you have to come up with a name that hasn’t been taken yet. For example, something that uses a unique string as a prefix. Your AWS_ACCESS_KEY (NOT YOUR SECRET KEY!) could work but I’ll leave it to your imagination to come up with something. I’ll just assume that you found an acceptable name.
The create_bucket method will create the requested bucket if it does not exist or will return the existing bucket if it does exist.
The example above assumes that you want to create a bucket in the standard US region. However, it is possible to create buckets in other locations. To do so, first import the Location object from the boto.s3.connection module, like this:
>>> from boto.s3.connection import Location >>> print '\n'.join(i for i in dir(Location) if i.isupper()) APNortheast APSoutheast APSoutheast2 DEFAULT EU SAEast USWest USWest2
As you can see, the Location object defines a number of possible locations. By default, the location is the empty string which is interpreted as the US Classic Region, the original S3 region. However, by specifying another location at the time the bucket is created, you can instruct S3 to create the bucket in that location. For example:
>>> conn.create_bucket('mybucket', location=Location.EU)
will create the bucket in the EU region (assuming the name is available).
Once you have a bucket, presumably you will want to store some data in it. S3 doesn’t care what kind of information you store in your objects or what format you use to store it. All you need is a key that is unique within your bucket.
The Key object is used in boto to keep track of data stored in S3. To store new data in S3, start by creating a new Key object:
>>> from boto.s3.key import Key >>> k = Key(bucket) >>> k.key = 'foobar' >>> k.set_contents_from_string('This is a test of S3')
The net effect of these statements is to create a new object in S3 with a key of “foobar” and a value of “This is a test of S3”. To validate that this worked, quit out of the interpreter and start it up again. Then:
>>> import boto >>> c = boto.connect_s3() >>> b = c.get_bucket('mybucket') # substitute your bucket name here >>> from boto.s3.key import Key >>> k = Key(b) >>> k.key = 'foobar' >>> k.get_contents_as_string() 'This is a test of S3'
So, we can definitely store and retrieve strings. A more interesting example may be to store the contents of a local file in S3 and then retrieve the contents to another local file.
>>> k = Key(b) >>> k.key = 'myfile' >>> k.set_contents_from_filename('foo.jpg') >>> k.get_contents_to_filename('bar.jpg')
There are a couple of things to note about this. When you send data to S3 from a file or filename, boto will attempt to determine the correct mime type for that file and send it as a Content-Type header. The boto package uses the standard mimetypes package in Python to do the mime type guessing. The other thing to note is that boto does stream the content to and from S3 so you should be able to send and receive large files without any problem.
When fetching a key that already exists, you have two options. If you’re uncertain whether a key exists (or if you need the metadata set on it, you can call Bucket.get_key(key_name_here). However, if you’re sure a key already exists within a bucket, you can skip the check for a key on the server.
>>> import boto >>> c = boto.connect_s3() >>> b = c.get_bucket('mybucket') # substitute your bucket name here # Will hit the API to check if it exists. >>> possible_key = b.get_key('mykey') # substitute your key name here # Won't hit the API. >>> key_we_know_is_there = b.get_key('mykey', validate=False)
At times the data you may want to store will be hundreds of megabytes or more in size. S3 allows you to split such files into smaller components. You upload each component in turn and then S3 combines them into the final object. While this is fairly straightforward, it requires a few extra steps to be taken. The example below makes use of the FileChunkIO module, so pip install FileChunkIO if it isn’t already installed.
>>> import math, os >>> import boto >>> from filechunkio import FileChunkIO # Connect to S3 >>> c = boto.connect_s3() >>> b = c.get_bucket('mybucket') # Get file info >>> source_path = 'path/to/your/file.ext' >>> source_size = os.stat(source_path).st_size # Create a multipart upload request >>> mp = b.initiate_multipart_upload(os.path.basename(source_path)) # Use a chunk size of 50 MiB (feel free to change this) >>> chunk_size = 52428800 >>> chunk_count = int(math.ceil(source_size / chunk_size)) # Send the file parts, using FileChunkIO to create a file-like object # that points to a certain byte range within the original file. We # set bytes to never exceed the original file size. >>> for i in range(chunk_count + 1): >>> offset = chunk_size * i >>> bytes = min(chunk_size, source_size - offset) >>> with FileChunkIO(source_path, 'r', offset=offset, bytes=bytes) as fp: >>> mp.upload_part_from_file(fp, part_num=i + 1) # Finish the upload >>> mp.complete_upload()
It is also possible to upload the parts in parallel using threads. The s3put script that ships with Boto provides an example of doing so using a thread pool.
Note that if you forget to call either mp.complete_upload() or mp.cancel_upload() you will be left with an incomplete upload and charged for the storage consumed by the uploaded parts. A call to bucket.get_all_multipart_uploads() can help to show lost multipart upload parts.
Once a bucket exists, you can access it by getting the bucket. For example:
>>> mybucket = conn.get_bucket('mybucket') # Substitute in your bucket name >>> mybucket.list() ...listing of keys in the bucket...
By default, this method tries to validate the bucket’s existence. You can override this behavior by passing validate=False.:
>>> nonexistent = conn.get_bucket('i-dont-exist-at-all', validate=False)
Changed in version 2.25.0.
If validate=False is passed, no request is made to the service (no charge/communication delay). This is only safe to do if you are sure the bucket exists.
If the default validate=True is passed, a request is made to the service to ensure the bucket exists. Prior to Boto v2.25.0, this fetched a list of keys (but with a max limit set to 0, always returning an empty list) in the bucket (& included better error messages), at an increased expense. As of Boto v2.25.0, this now performs a HEAD request (less expensive but worse error messages).
If you were relying on parsing the error message before, you should call something like:
bucket = conn.get_bucket('<bucket_name>', validate=False) bucket.get_all_keys(maxkeys=0)
If the bucket does not exist, a S3ResponseError will commonly be thrown. If you’d rather not deal with any exceptions, you can use the lookup method.:
>>> nonexistent = conn.lookup('i-dont-exist-at-all') >>> if nonexistent is None: ... print "No such bucket!" ... No such bucket!
Removing a bucket can be done using the delete_bucket method. For example:
>>> conn.delete_bucket('mybucket') # Substitute in your bucket name
The bucket must be empty of keys or this call will fail & an exception will be raised. You can remove a non-empty bucket by doing something like:
>>> full_bucket = conn.get_bucket('bucket-to-delete') # It's full of keys. Delete them all. >>> for key in full_bucket.list(): ... key.delete() ... # The bucket is empty now. Delete it. >>> conn.delete_bucket('bucket-to-delete')
This method can cause data loss! Be very careful when using it.
Additionally, be aware that using the above method for removing all keys and deleting the bucket involves a request for each key. As such, it’s not particularly fast & is very chatty.
In addition to accessing specific buckets via the create_bucket method you can also get a list of all available buckets that you have created.
>>> rs = conn.get_all_buckets()
This returns a ResultSet object (see the SQS Tutorial for more info on ResultSet objects). The ResultSet can be used as a sequence or list type object to retrieve Bucket objects.
>>> len(rs) 11 >>> for b in rs: ... print b.name ... <listing of available buckets> >>> b = rs
The S3 service provides the ability to control access to buckets and keys within s3 via the Access Control List (ACL) associated with each object in S3. There are two ways to set the ACL for an object:
To set a canned ACL for a bucket, use the set_acl method of the Bucket object. The argument passed to this method must be one of the four permissable canned policies named in the list CannedACLStrings contained in acl.py. For example, to make a bucket readable by anyone:
You can also set the ACL for Key objects, either by passing an additional argument to the above method:
>>> b.set_acl('public-read', 'foobar')
where ‘foobar’ is the key of some object within the bucket b or you can call the set_acl method of the Key object:
You can also retrieve the current ACL for a Bucket or Key object using the get_acl object. This method parses the AccessControlPolicy response sent by S3 and creates a set of Python objects that represent the ACL.
>>> acp = b.get_acl() >>> acp <boto.acl.Policy instance at 0x2e6940> >>> acp.acl <boto.acl.ACL instance at 0x2e69e0> >>> acp.acl.grants [<boto.acl.Grant instance at 0x2e6a08>] >>> for grant in acp.acl.grants: ... print grant.permission, grant.display_name, grant.email_address, grant.id ... FULL_CONTROL <boto.user.User instance at 0x2e6a30>
The Python objects representing the ACL can be found in the acl.py module of boto.
Both the Bucket object and the Key object also provide shortcut methods to simplify the process of granting individuals specific access. For example, if you want to grant an individual user READ access to a particular object in S3 you could do the following:
>>> key = b.lookup('mykeytoshare') >>> key.add_email_grant('READ', 'email@example.com')
The email address provided should be the one associated with the users AWS account. There is a similar method called add_user_grant that accepts the canonical id of the user rather than the email address.
S3 allows arbitrary user metadata to be assigned to objects within a bucket. To take advantage of this S3 feature, you should use the set_metadata and get_metadata methods of the Key object to set and retrieve metadata associated with an S3 object. For example:
>>> k = Key(b) >>> k.key = 'has_metadata' >>> k.set_metadata('meta1', 'This is the first metadata value') >>> k.set_metadata('meta2', 'This is the second metadata value') >>> k.set_contents_from_filename('foo.txt')
This code associates two metadata key/value pairs with the Key k. To retrieve those values later:
>>> k = b.get_key('has_metadata') >>> k.get_metadata('meta1') 'This is the first metadata value' >>> k.get_metadata('meta2') 'This is the second metadata value' >>>
Cross-origin resource sharing (CORS) defines a way for client web applications that are loaded in one domain to interact with resources in a different domain. With CORS support in Amazon S3, you can build rich client-side web applications with Amazon S3 and selectively allow cross-origin access to your Amazon S3 resources.
To create a CORS configuration and associate it with a bucket:
>>> from boto.s3.cors import CORSConfiguration >>> cors_cfg = CORSConfiguration() >>> cors_cfg.add_rule(['PUT', 'POST', 'DELETE'], 'https://www.example.com', allowed_header='*', max_age_seconds=3000, expose_header='x-amz-server-side-encryption') >>> cors_cfg.add_rule('GET', '*')
The above code creates a CORS configuration object with two rules.
To associate this configuration with a bucket:
>>> import boto >>> c = boto.connect_s3() >>> bucket = c.lookup('mybucket') >>> bucket.set_cors(cors_cfg)
To retrieve the CORS configuration associated with a bucket:
>>> cors_cfg = bucket.get_cors()
And, finally, to delete all CORS configurations from a bucket:
You can configure objects in S3 to transition to Glacier after a period of time. This is done using lifecycle policies. A lifecycle policy can also specify that an object should be deleted after a period of time. Lifecycle configurations are assigned to buckets and require these parameters:
For example, given a bucket s3-glacier-boto-demo, we can first retrieve the bucket:
>>> import boto >>> c = boto.connect_s3() >>> bucket = c.get_bucket('s3-glacier-boto-demo')
Then we can create a lifecycle object. In our example, we want all objects under logs/* to transition to Glacier 30 days after the object is created.
>>> from boto.s3.lifecycle import Lifecycle, Transition, Rule >>> to_glacier = Transition(days=30, storage_class='GLACIER') >>> rule = Rule('ruleid', 'logs/', 'Enabled', transition=to_glacier) >>> lifecycle = Lifecycle() >>> lifecycle.append(rule)
For API docs for the lifecycle objects, see boto.s3.lifecycle
We can now configure the bucket with this lifecycle policy:
>>> bucket.configure_lifecycle(lifecycle) True
You can also retrieve the current lifecycle policy for the bucket:
>>> current = bucket.get_lifecycle_config() >>> print current.transition <Transition: in: 30 days, GLACIER>
When an object transitions to Glacier, the storage class will be updated. This can be seen when you list the objects in a bucket:
>>> for key in bucket.list(): ... print key, key.storage_class ... <Key: s3-glacier-boto-demo,logs/testlog1.log> GLACIER
You can also use the prefix argument to the bucket.list method:
>>> print list(b.list(prefix='logs/testlog1.log')).storage_class u'GLACIER'
Once an object has been transitioned to Glacier, you can restore the object back to S3. To do so, you can use the boto.s3.key.Key.restore() method of the key object. The restore method takes an integer that specifies the number of days to keep the object in S3.
>>> import boto >>> c = boto.connect_s3() >>> bucket = c.get_bucket('s3-glacier-boto-demo') >>> key = bucket.get_key('logs/testlog1.log') >>> key.restore(days=5)
It takes about 4 hours for a restore operation to make a copy of the archive available for you to access. While the object is being restored, the ongoing_restore attribute will be set to True:
>>> key = bucket.get_key('logs/testlog1.log') >>> print key.ongoing_restore True
When the restore is finished, this value will be False and the expiry date of the object will be non None:
>>> key = bucket.get_key('logs/testlog1.log') >>> print key.ongoing_restore False >>> print key.expiry_date "Fri, 21 Dec 2012 00:00:00 GMT"
If there is no restore operation either in progress or completed, the ongoing_restore attribute will be None.
Once the object is restored you can then download the contents: