Recursive directory copy with Paramiko in Python

Question

I am new to Python scripting. I need to copy few folders from my local machine (windows) to Linux server. As of now, I am copying the folders by opening WinSCP console. I need to automate this process. I have written a below code in Python using Paramiko module library.

import paramiko
import os
transport = paramiko.Transport(('10.10.10.10', 22))
transport.connect(username='weblogic', password='weblogic')
sftp = paramiko.SFTPClient.from_transport(transport)
filepath = '/apps/logs'
localpath = 'C:\\Users\\Public\\test'
sftp.put(localpath,filepath)

Above is not working properly and giving below error. Can you please help me to copy the folder present in the windows path C:\Users\Public\test to Linux server path /apps/logs?

Traceback (most recent call last):
  File "C:\Users\Desktop\python\execute_script.py", line 28, in <module>
    sftp.put(localpath,filepath)
  File "C:\Python27\lib\paramiko\sftp_client.py", line 548, in put
    fl = file(localpath, 'rb')
IOError: [Errno 13] Permission denied: 'C:\\Users\\Public\\test'

Answer 1

Please check the below code from the link https://gist.github.com/johnfink8/2190472. I have used put_all method in the snippet.

import paramiko
import socket
import os
from stat import S_ISDIR

class SSHSession(object):
    # Usage:
    # Detects DSA or RSA from key_file, either as a string filename or a
    # file object.  Password auth is possible, but I will judge you for 
    # using it. So:
    # ssh=SSHSession('targetserver.com','root',key_file=open('mykey.pem','r'))
    # ssh=SSHSession('targetserver.com','root',key_file='/home/me/mykey.pem')
    # ssh=SSHSession('targetserver.com','root','mypassword')
    # ssh.put('filename','/remote/file/destination/path')
    # ssh.put_all('/path/to/local/source/dir','/path/to/remote/destination')
    # ssh.get_all('/path/to/remote/source/dir','/path/to/local/destination')
    # ssh.command('echo "Command to execute"')

    def __init__(self,hostname,username='root',key_file=None,password=None):
        #
        #  Accepts a file-like object (anything with a readlines() function)  
        #  in either dss_key or rsa_key with a private key.  Since I don't 
        #  ever intend to leave a server open to a password auth.
        #
        self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
        self.sock.connect((hostname,22))
        self.t = paramiko.Transport(self.sock)
        self.t.start_client()
        keys = paramiko.util.load_host_keys(os.path.expanduser('~/.ssh/known_hosts'))
        key = self.t.get_remote_server_key()
        # supposed to check for key in keys, but I don't much care right now to find the right notation
        if key_file is not None:
            if isinstance(key,str):
                key_file=open(key,'r')
            key_head=key_file.readline()
            key_file.seek(0)
            if 'DSA' in key_head:
                keytype=paramiko.DSSKey
            elif 'RSA' in key_head:
                keytype=paramiko.RSAKey
            else:
                raise Exception("Can't identify key type")
            pkey=keytype.from_private_key(key_file)
            self.t.auth_publickey(username, pkey)
        else:
            if password is not None:
                self.t.auth_password(username,password,fallback=False)
            else: raise Exception('Must supply either key_file or password')
        self.sftp=paramiko.SFTPClient.from_transport(self.t)

    def command(self,cmd):
        #  Breaks the command by lines, sends and receives 
        #  each line and its output separately
        #
        #  Returns the server response text as a string

        chan = self.t.open_session()
        chan.get_pty()
        chan.invoke_shell()
        chan.settimeout(20.0)
        ret=''
        try:
            ret+=chan.recv(1024)
        except:
            chan.send('\n')
            ret+=chan.recv(1024)
        for line in cmd.split('\n'):
            chan.send(line.strip() + '\n')
            ret+=chan.recv(1024)
        return ret

    def put(self,localfile,remotefile):
        #  Copy localfile to remotefile, overwriting or creating as needed.
        self.sftp.put(localfile,remotefile)

    def put_all(self,localpath,remotepath):
        #  recursively upload a full directory
        os.chdir(os.path.split(localpath)[0])
        parent=os.path.split(localpath)[1]
        for walker in os.walk(parent):
            try:
                self.sftp.mkdir(os.path.join(remotepath,walker[0]))
            except:
                pass
            for file in walker[2]:
                self.put(os.path.join(walker[0],file),os.path.join(remotepath,walker[0],file))

    def get(self,remotefile,localfile):
        #  Copy remotefile to localfile, overwriting or creating as needed.
        self.sftp.get(remotefile,localfile)

    def sftp_walk(self,remotepath):
        # Kindof a stripped down  version of os.walk, implemented for 
        # sftp.  Tried running it flat without the yields, but it really
        # chokes on big directories.
        path=remotepath
        files=[]
        folders=[]
        for f in self.sftp.listdir_attr(remotepath):
            if S_ISDIR(f.st_mode):
                folders.append(f.filename)
            else:
                files.append(f.filename)
        print (path,folders,files)
        yield path,folders,files
        for folder in folders:
            new_path=os.path.join(remotepath,folder)
            for x in self.sftp_walk(new_path):
                yield x

    def get_all(self,remotepath,localpath):
        #  recursively download a full directory
        #  Harder than it sounded at first, since paramiko won't walk
        #
        # For the record, something like this would gennerally be faster:
        # ssh user@host 'tar -cz /source/folder' | tar -xz

        self.sftp.chdir(os.path.split(remotepath)[0])
        parent=os.path.split(remotepath)[1]
        try:
            os.mkdir(localpath)
        except:
            pass
        for walker in self.sftp_walk(parent):
            try:
                os.mkdir(os.path.join(localpath,walker[0]))
            except:
                pass
            for file in walker[2]:
                self.get(os.path.join(walker[0],file),os.path.join(localpath,walker[0],file))
    def write_command(self,text,remotefile):
        #  Writes text to remotefile, and makes remotefile executable.
        #  This is perhaps a bit niche, but I was thinking I needed it.
        #  For the record, I was incorrect.
        self.sftp.open(remotefile,'w').write(text)
        self.sftp.chmod(remotefile,755)

Answer 2

In addition to the answer @user1041177, but here a way to do it when you are on windows to linux host (not really sure which kind of host actually).

I don't know why, but if I keep backslash onto remote path, I get a FileNotFoundException. The only way to work was to replace all '\' by '/'

Maybe someone could tell me the proper way to avoid this situation at all ?

Here a part of the exact same code above to give you breadcrumbs if you encounter the same issue :

def sftp_walk(socket, remotepath):
    remotepath = remotepath.replace('\\', '/')
    path = remotepath
    files = []
    folders = []
    for f in socket.listdir_attr(remotepath.replace('\\', '/')):
        if S_ISDIR(f.st_mode):
            folders.append(f.filename)
        else:
            files.append(f.filename)
    print(path, folders, files)
    yield path, folders, files
    for folder in folders:
        new_path = os.path.join(remotepath.replace('\\', '/'), folder)
        for x in sftp_walk(socket, new_path):
            yield x


def get_all(socket, remotepath, localpath):
    remotepath = remotepath.replace('\\', '/')
    socket.chdir(os.path.split(remotepath)[0])
    parent = os.path.split(remotepath)[1]
    try:
        os.mkdir(localpath)
    except:
        pass
    for walker in sftp_walk(socket, parent):
        try:
            os.mkdir(os.path.join(localpath, walker[0]).replace('\\', '/'))
        except:
            pass
        for file in walker[2]:
            socket.get(os.path.join(walker[0], file).replace('\\', '/'), os.path.join(localpath, walker[0], file).replace('\\', '/'))

BTW, I am not using those function inside an object, that's why their is 'socket' instead of 'self' because I call those function by passing the SFTP socket to them.

Finally have to say thank you to @user1041177, working like a charm.

Answer 3

I was trying to copy from a windows box to a linux box and got the same error as @Apex above. I was using the put_all method and I had to do some "replace" on parts of the code.

def put_all(self,localpath,remotepath):
        remotepath = remotepath.replace('\\', '/')
        #  recursively upload a full directory
        os.chdir(os.path.split(localpath)[0])
        parent=os.path.split(localpath)[1]
        for walker in os.walk(parent):
            try:
                self.sftp.mkdir(os.path.join(remotepath,walker[0]).replace('\\', '/'))
            except:
                pass
            for file in walker[2]:
                self.put(os.path.join(walker[0],file).replace('\\', '/'),os.path.join(remotepath,walker[0],file).replace('\\', '/'))

Answer 4

I found a few shortcomings with the above methods - first, the putter/getter doesn't function in the way you'd expect - if you want to put /foo/bar into /some/folder, you can't as it won't let you put files from a source folder to a different destination folder - the only thing you can do is put /foo/bar into /some/bar. In addition, you have to specify the source as /foo/bar and the destination as /some to end up with /some/bar - I find this confusing as it's not how most operating/ftp systems handle putting/getting/copying/etc. So, I improved on the answers listed above:

If you're going from Windows to Linux:

def put_dir(source, dest):
    source = os.path.expandvars(source).rstrip('\\').rstrip('/')
    dest = os.path.expandvars(dest).rstrip('\\').rstrip('/')

    for root, dirs, files in os.walk(source):
        for dir in dirs:
            try:
                sftp.mkdir(posixpath.join(dest, ''.join(root.rsplit(source))[1:].replace('\\', '/'), dir))
            except:
                pass
        for file in files:
            sftp.put(os.path.join(root, file), posixpath.join(dest, ''.join(root.rsplit(source))[1:].replace('\\', '/'), file))


source = '%USERPROFILE%\\Downloads\\'
dest = '/foo/bar'

put_dir(source, dest)

If you're just doing Windows then swap out posixpath.join with os.path.join and remove .replace('\\', '/'):

def put_dir(source, dest):
    source = os.path.expandvars(source).rstrip('\\').rstrip('/')
    dest = os.path.expandvars(dest).rstrip('\\').rstrip('/')

    for root, dirs, files in os.walk(source):
        for dir in dirs:
            try:
                sftp.mkdir(os.path.join(dest, ''.join(root.rsplit(source))[1:], dir))
            except:
                pass
        for file in files:
            sftp.put(os.path.join(root, file), os.path.join(dest, ''.join(root.rsplit(source))[1:], file))


source = '%USERPROFILE%\\Downloads\\'
dest = 'foo\\bar'

put_dir(source, dest)

The reason for the try statement is that sftp.mkdir errors out if the folder already exists.

Answer 5

Paramiko does not support recursive operations.

You can use pysftp. It's a wrapper around Paramiko that has more Python-ish look and feel and supports recursive operations. See

• pysftp.Connection.put_r()
• pysftp.Connection.get_r()

Or you can just base your code on pysftp source code. Or see my answer to Python pysftp get_r from Linux works fine on Linux but not on Windows.