Recursively PUT files to a remote server using FTP

Question

I'm currently in a situation where I have very limited access to a server, but need to upload and download a significant amount of files contained within a single directory structure. I don't have SSH access, so I can't use SCP - and rsync isn't an option either unfortunately.

I'm currently using ncftpput, which is great but seems to be quite slow (in spite of a fast connection).

Is there an alternative / better method I could look into?

(Please accept my apologies if this has been covered, I did a quick search prior to posting but didn't find anything that specifically answered my question)

Answer 1

Try using LFTP: http://lftp.yar.ru/

or YAFC: http://yafc.sourceforge.net/index.php

Answer 2

If you have a good connection, I would recommend mounting the ftp server via the GNOME or KDE file managers, or else using CurlFtpFS. Then you can treat it like just another folder.

Answer 3

I'm not familiar with ncftpput. For non-interactive FTP, I've always used the Perl Net::FTP module -- http://perldoc.perl.org/Net/FTP.html

This will be faster because you can login, then do all the transfers at once (it seems from a cursory glance that you execute ncftpput once for each file get/put).

Just remember to NEVER use ASCII mangling! This is the default, so use:

$ftp->binary

ASCII mangling needs to die in the same fire with MySQL automatic-timezone-interpreting.

Answer 4

Since I always end up having a problem with this, I'll post my notes here:

One thing I always get to confuse is the syntax; so below there is a bash tester script which creates some temporary directories, then starts a temporary ftp server, and compares rsync (in plain local file mode, as it doesn't support ftp) with lftp and ftpsync.

The thing is - you can use rsync /path/to/local /path/to/remote/, and rsync will automatically figure out, that you want a local subdirectory created under remote; however, for lftp or ftpsync you have to specify the target directory manually, as in ... /path/to/local /path/to/remote/local (if it doesn't exist it will be created).

You can find the ftpserver-cli.py in How do I temporarily run an FTP server? - Ask Ubuntu; and ftpsync is here: FTPsync (however, note it is buggy; see also Search/grep ftp remote filenames - Unix & Linux Stack Exchange);

Here is a shortened output of the puttest.sh script, showing the recursive put behavior in different cases:

$ bash puttest.sh 
Recreate directories; populate loctest, keep srvtest empty:
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt

*NOTE, rsync can automatically figure out parent dir:
+ rsync -a --exclude '*.git*' /tmp/loctest /tmp/srvtest/
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
└── loctest
    └── tempa1.txt
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt
cleanup:
+ rm -rf /tmp/srvtest/loctest

Start a temporary ftp server:
+ sudo bash -c 'python /path/to/pyftpdlib/ftpserver-cli.py --username=user --password=12345 --directory=/tmp/srvtest &'
+ sleep 1
Using: user: user pass: 12345 port: 21 dir: /tmp/srvtest
[I 14-03-02 23:24:01] >>> starting FTP server on 127.0.0.1:21, pid=21549 <<<
[I 14-03-02 23:24:01] poller: <class 'pyftpdlib.ioloop.Epoll'>
[I 14-03-02 23:24:01] masquerade (NAT) address: None
[I 14-03-02 23:24:01] passive ports: None
[I 14-03-02 23:24:01] use sendfile(2): False
test with lftp:

*NOTE, lftp syncs *contents* of local dir (rsync-like syntax doesn't create target dir):
+ lftp -e 'mirror -R -x ".*\.git.*" /tmp/loctest / ; exit' -u user,12345 127.0.0.1
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
└── tempa1.txt
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt
cleanup:
+ rm -rf /tmp/srvtest/tempa1.txt

*NOTE, specify lftp target dir explicitly (will be autocreated):
+ lftp -e 'mirror -R -x ".*\.git.*" /tmp/loctest /loctest ; exit' -u user,12345 127.0.0.1
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
└── loctest
    └── tempa1.txt
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt
cleanup:
+ sudo rm -rf /tmp/srvtest/loctest

*NOTE, ftpsync syncs *contents* of local dir (rsync-like syntax doesn't create target dir); also info mode -i is buggy (it puts, although it shouldn't):

*NOTE, ftpsync --ignoremask is for older unused code; use --exclude instead (but it is buggy; need to change  in source)
+ /path/to/ftpsync/ftpsync -i -d '--exclude=.*\.git.*' /tmp/loctest ftp://user:12345@127.0.0.1/
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
└── tempa1.txt
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt
cleanup:
+ sudo rm -rf /tmp/srvtest/tempa1.txt

*NOTE, specify ftpsync target dir explicitly (will be autocreated):
+ /path/to/ftpsync/ftpsync -i -d '--exclude=.*\.git.*' /tmp/loctest ftp://user:12345@127.0.0.1/loctest
show dirs:
+ tree --noreport -a /tmp/srvtest /tmp/loctest
/tmp/srvtest
└── loctest
    └── tempa1.txt
/tmp/loctest
├── .git
│   └── tempa2.txt
└── tempa1.txt
cleanup:
+ sudo rm -rf /tmp/srvtest/loctest
+ sudo pkill -f ftpserver-cli.py

And, here is the puttest.sh script:

#!/usr/bin/env bash
set -x

# change these to match your installations:
FTPSRVCLIPATH="/path/to/pyftpdlib"
FTPSYNCPATH="/path/to/ftpsync"

{ echo "Recreate directories; populate loctest, keep srvtest empty:"; } 2>/dev/null

sudo rm -rf /tmp/srvtest /tmp/loctest

mkdir /tmp/srvtest

mkdir -p /tmp/loctest/.git
echo aaa > /tmp/loctest/tempa1.txt
echo aaa > /tmp/loctest/.git/tempa2.txt

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo -e "\n*NOTE, rsync can automatically figure out parent dir:"; } 2>/dev/null

rsync -a --exclude '*.git*' /tmp/loctest /tmp/srvtest/

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo "cleanup:"; } 2>/dev/null
rm -rf /tmp/srvtest/*

{ echo -e "\nStart a temporary ftp server:"; } 2>/dev/null

# https://askubuntu.com/questions/17084/how-do-i-temporarily-run-an-ftp-server

sudo bash -c "python $FTPSRVCLIPATH/ftpserver-cli.py --username=user --password=12345 --directory=/tmp/srvtest &"
sleep 1

{ echo "test with lftp:"; } 2>/dev/null
# see http://russbrooks.com/2010/11/19/lftp-cheetsheet
# The -R switch means "reverse mirror" which means "put" [upload].
{ echo -e "\n*NOTE, lftp syncs *contents* of local dir (rsync-like syntax doesn't create target dir):"; } 2>/dev/null

lftp -e 'mirror -R -x ".*\.git.*" /tmp/loctest / ; exit' -u user,12345 127.0.0.1

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo "cleanup:"; } 2>/dev/null
rm -rf /tmp/srvtest/*

{ echo -e "\n*NOTE, specify lftp target dir explicitly (will be autocreated):"; } 2>/dev/null

lftp -e 'mirror -R -x ".*\.git.*" /tmp/loctest /loctest ; exit' -u user,12345 127.0.0.1

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo "cleanup:"; } 2>/dev/null
sudo rm -rf /tmp/srvtest/*

{ echo -e "\n*NOTE, ftpsync syncs *contents* of local dir (rsync-like syntax doesn't create target dir); also info mode -i is buggy (it puts, although it shouldn't):"; } 2>/dev/null
{ echo -e "\n*NOTE, ftpsync --ignoremask is for older unused code; use --exclude instead (but it is buggy; need to change `  'exclude=s' => \$opt::exclude,` in source)"; } 2>/dev/null

$FTPSYNCPATH/ftpsync -i -d --exclude='.*\.git.*' /tmp/loctest ftp://user:12345@127.0.0.1/

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo "cleanup:"; } 2>/dev/null
sudo rm -rf /tmp/srvtest/*

{ echo -e "\n*NOTE, specify ftpsync target dir explicitly (will be autocreated):"; } 2>/dev/null

$FTPSYNCPATH/ftpsync -i -d --exclude='.*\.git.*' /tmp/loctest ftp://user:12345@127.0.0.1/loctest

{ echo "show dirs:"; } 2>/dev/null
tree --noreport -a /tmp/srvtest /tmp/loctest

{ echo "cleanup:"; } 2>/dev/null
sudo rm -rf /tmp/srvtest/*


sudo pkill -f ftpserver-cli.py

{ set +x; } 2>/dev/null