python - Boto3 to download all files from a S3 Bucket

Question

I'm using boto3 to get files from s3 bucket. I need a similar functionality like aws s3 sync

My current code is

#!/usr/bin/python
import boto3
s3=boto3.client('s3')
list=s3.list_objects(Bucket='my_bucket_name')['Contents']
for key in list:
    s3.download_file('my_bucket_name', key['Key'], key['Key'])

This is working fine, as long as the bucket has only files. If a folder is present inside the bucket, its throwing an error

Traceback (most recent call last):
  File "./test", line 6, in <module>
    s3.download_file('my_bucket_name', key['Key'], key['Key'])
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/inject.py", line 58, in download_file
    extra_args=ExtraArgs, callback=Callback)
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/transfer.py", line 651, in download_file
    extra_args, callback)
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/transfer.py", line 666, in _download_file
    self._get_object(bucket, key, filename, extra_args, callback)
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/transfer.py", line 690, in _get_object
    extra_args, callback)
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/transfer.py", line 707, in _do_get_object
    with self._osutil.open(filename, 'wb') as f:
  File "/usr/local/lib/python2.7/dist-packages/boto3/s3/transfer.py", line 323, in open
    return open(filename, mode)
IOError: [Errno 2] No such file or directory: 'my_folder/.8Df54234'

Is this a proper way to download a complete s3 bucket using boto3. How to download folders.

Answer 1

When working with buckets that have 1000+ objects its necessary to implement a solution that uses the NextContinuationToken on sequential sets of, at most, 1000 keys. This solution first compiles a list of objects then iteratively creates the specified directories and downloads the existing objects.

import boto3
import os

s3_client = boto3.client('s3')

def download_dir(prefix, local, bucket, client=s3_client):
    """
    params:
    - prefix: pattern to match in s3
    - local: local path to folder in which to place files
    - bucket: s3 bucket with target contents
    - client: initialized s3 client object
    """
    keys = []
    dirs = []
    next_token = ''
    base_kwargs = {
        'Bucket':bucket,
        'Prefix':prefix,
    }
    while next_token is not None:
        kwargs = base_kwargs.copy()
        if next_token != '':
            kwargs.update({'ContinuationToken': next_token})
        results = client.list_objects_v2(**kwargs)
        contents = results.get('Contents')
        for i in contents:
            k = i.get('Key')
            if k[-1] != '/':
                keys.append(k)
            else:
                dirs.append(k)
        next_token = results.get('NextContinuationToken')
    for d in dirs:
        dest_pathname = os.path.join(local, d)
        if not os.path.exists(os.path.dirname(dest_pathname)):
            os.makedirs(os.path.dirname(dest_pathname))
    for k in keys:
        dest_pathname = os.path.join(local, k)
        if not os.path.exists(os.path.dirname(dest_pathname)):
            os.makedirs(os.path.dirname(dest_pathname))
        client.download_file(bucket, k, dest_pathname)

Answer 2

I have the same needs and created the following function that download recursively the files.

The directories are created locally only if they contain files.

import boto3
import os

def download_dir(client, resource, dist, local='/tmp', bucket='your_bucket'):
    paginator = client.get_paginator('list_objects')
    for result in paginator.paginate(Bucket=bucket, Delimiter='/', Prefix=dist):
        if result.get('CommonPrefixes') is not None:
            for subdir in result.get('CommonPrefixes'):
                download_dir(client, resource, subdir.get('Prefix'), local, bucket)
        for file in result.get('Contents', []):
            dest_pathname = os.path.join(local, file.get('Key'))
            if not os.path.exists(os.path.dirname(dest_pathname)):
                os.makedirs(os.path.dirname(dest_pathname))
            resource.meta.client.download_file(bucket, file.get('Key'), dest_pathname)

The function is called that way:

def _start():
    client = boto3.client('s3')
    resource = boto3.resource('s3')
    download_dir(client, resource, 'clientconf/', '/tmp', bucket='my-bucket')

Answer 3

Amazon S3 does not have folders/directories. It is a flat file structure.

To maintain the appearance of directories, path names are stored as part of the object Key (filename). For example:

• images/foo.jpg

In this case, the whole Key is images/foo.jpg, rather than just foo.jpg.

I suspect that your problem is that boto is returning a file called my_folder/.8Df54234 and is attempting to save it to the local filesystem. However, your local filesystem interprets the my_folder/ portion as a directory name, and that directory does not exist on your local filesystem.

You could either truncate the filename to only save the .8Df54234 portion, or you would have to create the necessary directories before writing files. Note that it could be multi-level nested directories.

An easier way would be to use the AWS Command-Line Interface (CLI), which will do all this work for you, eg:

aws s3 cp --recursive s3://my_bucket_name local_folder

There's also a sync option that will only copy new and modified files.

Answer 4

import os
import boto3

#initiate s3 resource
s3 = boto3.resource('s3')

# select bucket
my_bucket = s3.Bucket('my_bucket_name')

# download file into current directory
for s3_object in my_bucket.objects.all():
    # Need to split s3_object.key into path and file name, else it will give error file not found.
    path, filename = os.path.split(s3_object.key)
    my_bucket.download_file(s3_object.key, filename)

Answer 5

I'm currently achieving the task, by using the following

#!/usr/bin/python
import boto3
s3=boto3.client('s3')
list=s3.list_objects(Bucket='bucket')['Contents']
for s3_key in list:
    s3_object = s3_key['Key']
    if not s3_object.endswith("/"):
        s3.download_file('bucket', s3_object, s3_object)
    else:
        import os
        if not os.path.exists(s3_object):
            os.makedirs(s3_object)

Although, it does the job, I'm not sure its good to do this way. I'm leaving it here to help other users and further answers, with better manner of achieving this

Answer 6

Better late than never:) The previous answer with paginator is really good. However it is recursive, and you might end up hitting Python's recursion limits. Here's an alternate approach, with a couple of extra checks.

import os
import errno
import boto3


def assert_dir_exists(path):
    """
    Checks if directory tree in path exists. If not it created them.
    :param path: the path to check if it exists
    """
    try:
        os.makedirs(path)
    except OSError as e:
        if e.errno != errno.EEXIST:
            raise


def download_dir(client, bucket, path, target):
    """
    Downloads recursively the given S3 path to the target directory.
    :param client: S3 client to use.
    :param bucket: the name of the bucket to download from
    :param path: The S3 directory to download.
    :param target: the local directory to download the files to.
    """

    # Handle missing / at end of prefix
    if not path.endswith('/'):
        path += '/'

    paginator = client.get_paginator('list_objects_v2')
    for result in paginator.paginate(Bucket=bucket, Prefix=path):
        # Download each file individually
        for key in result['Contents']:
            # Calculate relative path
            rel_path = key['Key'][len(path):]
            # Skip paths ending in /
            if not key['Key'].endswith('/'):
                local_file_path = os.path.join(target, rel_path)
                # Make sure directories exist
                local_file_dir = os.path.dirname(local_file_path)
                assert_dir_exists(local_file_dir)
                client.download_file(bucket, key['Key'], local_file_path)


client = boto3.client('s3')

download_dir(client, 'bucket-name', 'path/to/data', 'downloads')

Answer 7

I have a workaround for this that runs the AWS CLI in the same process.

Install awscli as python lib:

pip install awscli

Then define this function:

from awscli.clidriver import create_clidriver

def aws_cli(*cmd):
    old_env = dict(os.environ)
    try:

        # Environment
        env = os.environ.copy()
        env['LC_CTYPE'] = u'en_US.UTF'
        os.environ.update(env)

        # Run awscli in the same process
        exit_code = create_clidriver().main(*cmd)

        # Deal with problems
        if exit_code > 0:
            raise RuntimeError('AWS CLI exited with code {}'.format(exit_code))
    finally:
        os.environ.clear()
        os.environ.update(old_env)

To execute:

aws_cli('s3', 'sync', '/path/to/source', 's3://bucket/destination', '--delete')

Answer 8

It is a very bad idea to get all files in one go, you should rather get it in batches.

One implementation which I use to fetch a particular folder (directory) from S3 is,

def get_directory(directory_path, download_path, exclude_file_names):
    # prepare session
    session = Session(aws_access_key_id, aws_secret_access_key, region_name)

    # get instances for resource and bucket
    resource = session.resource('s3')
    bucket = resource.Bucket(bucket_name)

    for s3_key in self.client.list_objects(Bucket=self.bucket_name, Prefix=directory_path)['Contents']:
        s3_object = s3_key['Key']
        if s3_object not in exclude_file_names:
            bucket.download_file(file_path, download_path + str(s3_object.split('/')[-1])

and still if you want to get the whole bucket use it via CIL as @John Rotenstein mentioned as below,

aws s3 cp --recursive s3://bucket_name download_path

Answer 9

From AWS S3 Docs (How do I use folders in an S3 bucket?):

In Amazon S3, buckets and objects are the primary resources, and objects are stored in buckets. Amazon S3 has a flat structure instead of a hierarchy like you would see in a file system. However, for the sake of organizational simplicity, the Amazon S3 console supports the folder concept as a means of grouping objects. Amazon S3 does this by using a shared name prefix for objects (that is, objects have names that begin with a common string). Object names are also referred to as key names.

For example, you can create a folder on the console named photos and store an object named myphoto.jpg in it. The object is then stored with the key name photos/myphoto.jpg, where photos/ is the prefix.

To download all files from "mybucket" into the current directory respecting the bucket's emulated directory structure (creating the folders from the bucket if they don't already exist locally):

import boto3
import os

bucket_name = "mybucket"
s3 = boto3.client("s3")
objects = s3.list_objects(Bucket = bucket_name)["Contents"]
for s3_object in objects:
    s3_key = s3_object["Key"]
    path, filename = os.path.split(s3_key)
    if len(path) != 0 and not os.path.exists(path):
        os.makedirs(path)
    if not s3_key.endswith("/"):
        download_to = path + '/' + filename if path else filename
        s3.download_file(bucket_name, s3_key, download_to)

Answer 10

import boto3, os

s3 = boto3.client('s3')

def download_bucket(bucket):
    paginator = s3.get_paginator('list_objects_v2')
    pages = paginator.paginate(Bucket=bucket)
    for page in pages:
      if 'Contents' in page:
        for obj in page['Contents']:
            os.path.dirname(obj['Key']) and os.makedirs(os.path.dirname(obj['Key']), exist_ok=True) 
            try:
                s3.download_file(bucket, obj['Key'], obj['Key'])
            except NotADirectoryError:
                pass

# Change bucket_name to name of bucket that you want to download
download_bucket(bucket_name)

This should work for all number of objects (also when there are more than 1000). Each paginator page can contain up to 1000 objects.Notice extra param in os.makedirs function - exist_ok=True which cause that it's not throwing error when path exist)

Answer 11

I've updated Grant's answer to run in parallel, it's much faster if anyone is interested:

from concurrent import futures
import os
import boto3

def download_dir(prefix, local, bucket):

    client = boto3.client('s3')

    def create_folder_and_download_file(k):
        dest_pathname = os.path.join(local, k)
        if not os.path.exists(os.path.dirname(dest_pathname)):
            os.makedirs(os.path.dirname(dest_pathname))
        print(f'downloading {k} to {dest_pathname}')
        client.download_file(bucket, k, dest_pathname)

    keys = []
    dirs = []
    next_token = ''
    base_kwargs = {
        'Bucket': bucket,
        'Prefix': prefix,
    }
    while next_token is not None:
        kwargs = base_kwargs.copy()
        if next_token != '':
            kwargs.update({'ContinuationToken': next_token})
        results = client.list_objects_v2(**kwargs)
        contents = results.get('Contents')
        for i in contents:
            k = i.get('Key')
            if k[-1] != '/':
                keys.append(k)
            else:
                dirs.append(k)
        next_token = results.get('NextContinuationToken')
    for d in dirs:
        dest_pathname = os.path.join(local, d)
        if not os.path.exists(os.path.dirname(dest_pathname)):
            os.makedirs(os.path.dirname(dest_pathname))
    with futures.ThreadPoolExecutor() as executor:
        futures.wait(
            [executor.submit(create_folder_and_download_file, k) for k in keys],
            return_when=futures.FIRST_EXCEPTION,
        )

Answer 12

for objs in my_bucket.objects.all():
    print(objs.key)
    path='/tmp/'+os.sep.join(objs.key.split(os.sep)[:-1])
    try:
        if not os.path.exists(path):
            os.makedirs(path)
        my_bucket.download_file(objs.key, '/tmp/'+objs.key)
    except FileExistsError as fe:                          
        print(objs.key+' exists')

This code will download the content in /tmp/ directory. If you want you can change the directory.

Answer 13

If you want to call a bash script using python, here is a simple method to load a file from a folder in S3 bucket to a local folder (in a Linux machine) :

import boto3
import subprocess
import os

###TOEDIT###
my_bucket_name = "your_my_bucket_name"
bucket_folder_name = "your_bucket_folder_name"
local_folder_path = "your_local_folder_path"
###TOEDIT###

# 1.Load thes list of files existing in the bucket folder
FILES_NAMES = []
s3 = boto3.resource('s3')
my_bucket = s3.Bucket('{}'.format(my_bucket_name))
for object_summary in my_bucket.objects.filter(Prefix="{}/".format(bucket_folder_name)):
#     print(object_summary.key)
    FILES_NAMES.append(object_summary.key)

# 2.List only new files that do not exist in local folder (to not copy everything!)
new_filenames = list(set(FILES_NAMES )-set(os.listdir(local_folder_path)))

# 3.Time to load files in your destination folder 
for new_filename in new_filenames:
    upload_S3files_CMD = """aws s3 cp s3://{}/{}/{} {}""".format(my_bucket_name,bucket_folder_name,new_filename ,local_folder_path)

    subprocess_call = subprocess.call([upload_S3files_CMD], shell=True)
    if subprocess_call != 0:
        print("ALERT: loading files not working correctly, please re-check new loaded files")

Answer 14

I got the similar requirement and got help from reading few of the above solutions and across other websites, I have came up with below script, Just wanted to share if it might help anyone.

from boto3.session import Session
import os

def sync_s3_folder(access_key_id,secret_access_key,bucket_name,folder,destination_path):    
    session = Session(aws_access_key_id=access_key_id,aws_secret_access_key=secret_access_key)
    s3 = session.resource('s3')
    your_bucket = s3.Bucket(bucket_name)
    for s3_file in your_bucket.objects.all():
        if folder in s3_file.key:
            file=os.path.join(destination_path,s3_file.key.replace('/','\\'))
            if not os.path.exists(os.path.dirname(file)):
                os.makedirs(os.path.dirname(file))
            your_bucket.download_file(s3_file.key,file)
sync_s3_folder(access_key_id,secret_access_key,bucket_name,folder,destination_path)

Answer 15

Reposting @glefait 's answer with an if condition at the end to avoid os error 20. The first key it gets is the folder name itself which cannot be written in the destination path.

def download_dir(client, resource, dist, local='/tmp', bucket='your_bucket'):
    paginator = client.get_paginator('list_objects')
    for result in paginator.paginate(Bucket=bucket, Delimiter='/', Prefix=dist):
        if result.get('CommonPrefixes') is not None:
            for subdir in result.get('CommonPrefixes'):
                download_dir(client, resource, subdir.get('Prefix'), local, bucket)
        for file in result.get('Contents', []):
            print("Content: ",result)
            dest_pathname = os.path.join(local, file.get('Key'))
            print("Dest path: ",dest_pathname)
            if not os.path.exists(os.path.dirname(dest_pathname)):
                print("here last if")
                os.makedirs(os.path.dirname(dest_pathname))
            print("else file key: ", file.get('Key'))
            if not file.get('Key') == dist:
                print("Key not equal? ",file.get('Key'))
                resource.meta.client.download_file(bucket, file.get('Key'), dest_pathname)enter code here

Answer 16

I have been running into this problem for a while and with all of the different forums I've been through I haven't see a full end-to-end snip-it of what works. So, I went ahead and took all the pieces (add some stuff on my own) and have created a full end-to-end S3 Downloader!

This will not only download files automatically but if the S3 files are in subdirectories, it will create them on the local storage. In my application's instance, I need to set permissions and owners so I have added that too (can be comment out if not needed).

This has been tested and works in a Docker environment (K8) but I have added the environmental variables in the script just in case you want to test/run it locally.

I hope this helps someone out in their quest of finding S3 Download automation. I also welcome any advice, info, etc. on how this can be better optimized if needed.

#!/usr/bin/python3
import gc
import logging
import os
import signal
import sys
import time
from datetime import datetime

import boto
from boto.exception import S3ResponseError
from pythonjsonlogger import jsonlogger

formatter = jsonlogger.JsonFormatter('%(message)%(levelname)%(name)%(asctime)%(filename)%(lineno)%(funcName)')

json_handler_out = logging.StreamHandler()
json_handler_out.setFormatter(formatter)

#Manual Testing Variables If Needed
#os.environ["DOWNLOAD_LOCATION_PATH"] = "some_path"
#os.environ["BUCKET_NAME"] = "some_bucket"
#os.environ["AWS_ACCESS_KEY"] = "some_access_key"
#os.environ["AWS_SECRET_KEY"] = "some_secret"
#os.environ["LOG_LEVEL_SELECTOR"] = "DEBUG, INFO, or ERROR"

#Setting Log Level Test
logger = logging.getLogger('json')
logger.addHandler(json_handler_out)
logger_levels = {
    'ERROR' : logging.ERROR,
    'INFO' : logging.INFO,
    'DEBUG' : logging.DEBUG
}
logger_level_selector = os.environ["LOG_LEVEL_SELECTOR"]
logger.setLevel(logger_level_selector)

#Getting Date/Time
now = datetime.now()
logger.info("Current date and time : ")
logger.info(now.strftime("%Y-%m-%d %H:%M:%S"))

#Establishing S3 Variables and Download Location
download_location_path = os.environ["DOWNLOAD_LOCATION_PATH"]
bucket_name = os.environ["BUCKET_NAME"]
aws_access_key_id = os.environ["AWS_ACCESS_KEY"]
aws_access_secret_key = os.environ["AWS_SECRET_KEY"]
logger.debug("Bucket: %s" % bucket_name)
logger.debug("Key: %s" % aws_access_key_id)
logger.debug("Secret: %s" % aws_access_secret_key)
logger.debug("Download location path: %s" % download_location_path)

#Creating Download Directory
if not os.path.exists(download_location_path):
    logger.info("Making download directory")
    os.makedirs(download_location_path)

#Signal Hooks are fun
class GracefulKiller:
    kill_now = False
    def __init__(self):
        signal.signal(signal.SIGINT, self.exit_gracefully)
        signal.signal(signal.SIGTERM, self.exit_gracefully)
    def exit_gracefully(self, signum, frame):
        self.kill_now = True

#Downloading from S3 Bucket
def download_s3_bucket():
    conn = boto.connect_s3(aws_access_key_id, aws_access_secret_key)
    logger.debug("Connection established: ")
    bucket = conn.get_bucket(bucket_name)
    logger.debug("Bucket: %s" % str(bucket))
    bucket_list = bucket.list()
#    logger.info("Number of items to download: {0}".format(len(bucket_list)))

    for s3_item in bucket_list:
        key_string = str(s3_item.key)
        logger.debug("S3 Bucket Item to download: %s" % key_string)
        s3_path = download_location_path + "/" + key_string
        logger.debug("Downloading to: %s" % s3_path)
        local_dir = os.path.dirname(s3_path)

        if not os.path.exists(local_dir):
            logger.info("Local directory doesn't exist, creating it... %s" % local_dir)
            os.makedirs(local_dir)
            logger.info("Updating local directory permissions to %s" % local_dir)
#Comment or Uncomment Permissions based on Local Usage
            os.chmod(local_dir, 0o775)
            os.chown(local_dir, 60001, 60001)
        logger.debug("Local directory for download: %s" % local_dir)
        try:
            logger.info("Downloading File: %s" % key_string)
            s3_item.get_contents_to_filename(s3_path)
            logger.info("Successfully downloaded File: %s" % s3_path)
            #Updating Permissions
            logger.info("Updating Permissions for %s" % str(s3_path))
#Comment or Uncomment Permissions based on Local Usage
            os.chmod(s3_path, 0o664)
            os.chown(s3_path, 60001, 60001)
        except (OSError, S3ResponseError) as e:
            logger.error("Fatal error in s3_item.get_contents_to_filename", exc_info=True)
            # logger.error("Exception in file download from S3: {}".format(e))
            continue
        logger.info("Deleting %s from S3 Bucket" % str(s3_item.key))
        s3_item.delete()

def main():
    killer = GracefulKiller()
    while not killer.kill_now:
        logger.info("Checking for new files on S3 to download...")
        download_s3_bucket()
        logger.info("Done checking for new files, will check in 120s...")
        gc.collect()
        sys.stdout.flush()
        time.sleep(120)
if __name__ == '__main__':
    main()

Answer 17

A lot of the solutions here get pretty complicated. If you're looking for something simpler, cloudpathlib wraps things in a nice way for this use case that will download directories or files.

from cloudpathlib import CloudPath

cp = CloudPath("s3://bucket/product/myproject/2021-02-15/")
cp.download_to("local_folder")

Note: for large folders with lots of files, awscli at the command line is likely faster.

Answer 18

Yet another parallel downloader using asyncio/aioboto

import os, time
import asyncio
from itertools import chain
import json
from typing import List
from json.decoder import WHITESPACE
import logging
from functools import partial
from pprint import pprint as pp

# Third Party
import asyncpool
import aiobotocore.session
import aiobotocore.config

_NUM_WORKERS = 50


bucket_name= 'test-data'
bucket_prefix= 'etl2/test/20210330/f_api'


async def save_to_file(s3_client, bucket: str, key: str):
    
    response = await s3_client.get_object(Bucket=bucket, Key=key)
    async with response['Body'] as stream:
        content = await stream.read()
    
    if 1:
        fn =f'out/downloaded/{bucket_name}/{key}'

        dn= os.path.dirname(fn)
        if not isdir(dn):
            os.makedirs(dn,exist_ok=True)
        if 1:
            with open(fn, 'wb') as fh:
                fh.write(content)
                print(f'Downloaded to: {fn}')
   
    return [0]


async def go(bucket: str, prefix: str) -> List[dict]:
    """
    Returns list of dicts of object contents

    :param bucket: s3 bucket
    :param prefix: s3 bucket prefix
    :return: list of download statuses
    """
    logging.basicConfig(level=logging.INFO)
    logger = logging.getLogger()

    session = aiobotocore.session.AioSession()
    config = aiobotocore.config.AioConfig(max_pool_connections=_NUM_WORKERS)
    contents = []
    async with session.create_client('s3', config=config) as client:
        worker_co = partial(save_to_file, client, bucket)
        async with asyncpool.AsyncPool(None, _NUM_WORKERS, 's3_work_queue', logger, worker_co,
                                       return_futures=True, raise_on_join=True, log_every_n=10) as work_pool:
            # list s3 objects using paginator
            paginator = client.get_paginator('list_objects')
            async for result in paginator.paginate(Bucket=bucket, Prefix=prefix):
                for c in result.get('Contents', []):
                    contents.append(await work_pool.push(c['Key'], client))

    # retrieve results from futures
    contents = [c.result() for c in contents]
    return list(chain.from_iterable(contents))


def S3_download_bucket_files():
    s = time.perf_counter()
    _loop = asyncio.get_event_loop()
    _result = _loop.run_until_complete(go(bucket_name, bucket_prefix))
    assert sum(_result)==0, _result
    print(_result)
    elapsed = time.perf_counter() - s
    print(f"{__file__} executed in {elapsed:0.2f} seconds.")

It will fetch list of files from S3 first and then download using aioboto with _NUM_WORKERS=50 reading data in parallel from the network.