NAME
ncftpspooler - Global batch FTP job processor daemon
SYNOPSIS
ncftpspooler -d [options]
ncftpspooler -l [options]
OPTIONS
Command line flags:
-d Begin background processing of FTP jobs in the designated FTP
job queue directory.
-q XX Use this option to specify a directory to use as the FTP job
queue instead of the default directory, /var/spool/ncftp.
-o XX Use this option to specify a filename to use as the log file.
By default, (and rather inappropriately) the program simply
uses a file called log in the job queue directory. If you
don’t want a log, use this option to specify /dev/null.
-l Lists the contents of the job queue directory.
-s XX When the job queue is empty, the program sleeps 120 seconds and
then checks again to see if a new job has been submitted. Use
this option to change the number of seconds used for this
delay.
DESCRIPTION
The ncftpspooler program evolved from the ncftpbatch program. The
ncftpbatch program was originally designed as a ‘‘personal FTP
spooler’’ which would process a single background job a particular user
and exit when it finished; the ncftpspooler program is a ‘‘global FTP
spooler’’ which stays running and processes background jobs as they are
submitted.
The job queue directory is monitored for specially-named and formatted
text files. Each file serves as a single FTP job. The name of the job
file contains the type of FTP job (get or put), a timestamp indicating
the earliest the job should be processed, and optionally some
additional information to make it easier to create unique job files
(i.e. a sequence number). The contents of the job files have
information such as the remote server machine to FTP to, username,
password, remote pathname, etc.
Your job queue directory must be readable and writable by the user that
you plan to run ncftpspooler as, so that jobs can be removed or renamed
within the queue.
More importantly, the user that is running the program will need
adequate privileges to access the local files that are involved in the
FTPing. I.e., if your spooler is going to be processing jobs which
upload files to remote servers, then the user will need read permission
on the local files that will be uploaded (and directory access
permission the parent directories). Likewise, if your spooler is going
to be processing jobs which download files, then the user would need to
be able to write to the local directories.
Once you have created your spool directory with appropriate permissions
and ownerships, you can run ncftpspooler -d to launch the spooler
daemon. You can run additional spoolers if you want to process more
than FTP job from the same job queue directory simultaneously. You can
then monitor the log file (i.e., using tail -f ) to track the progress
of the spooler. Most of the time it won’t be doing anything, unless
job files have appeared in the job queue directory.
JOB FILE NAMES
When the ncftpspooler program monitors the job queue directory, it
ignores any files that do not follow the naming convention for job
files. The job files must be prefixed in the format of
X-YYYYMMDD-hhmmss where X denotes a job type, YYYY is the four-digit
year, MM is the two-digit month number, DD is the two-digit day of the
month, hh is the two-digit hour of the day (00-23), mm is the two-digit
minute, and ss is the two-digit second. The date and time represent
the earliest time you want the job to be run.
The job type can be g for a get (download from remote host), or p for
aput (upload to remote host).
As an example, if you wanted to schedule an upload to occur at 11:45 PM
on December 7, 2001, a job file could be named
p-20011207-234500
In practice, the job files include additional information such as a
sequence number or process ID. This makes it easier to create unique
job file names. Here is the same example, with a process ID and a
sequence number:
p-20011207-234500-1234-2
When submitting job files to the queue directory, be sure to use a dash
character after the hhmmss field if you choose to append any additional
data to the job file name.
JOB FILE CONTENTS
Job files are ordinary text files, so that they can be created by hand.
Each line of the file is a key-pair in the format variable=value, or is
a comment line beginning with an octothorpe character (#), or is a
blank line. Here is an example job file:
# This is a NcFTP spool file entry.
job-name=g-20011016-100656-008299-1
op=get
hostname=ftp.freebsd.org
xtype=I
passive=1
remote-dir=pub/FreeBSD
local-dir=/tmp
remote-file=README.TXT
local-file=readme.txt
Job files are flexible since they follow an easy-to-use format and do
not have many requirements, but there are a few mandatory parameters
that must appear for the spooler to be able to process the job.
op The operation (job type) to perform. Valid values are get and
put.
hostname
The remote host to FTP to. This may be an IP address or a DNS
name (i.e. ftp.example.com).
For a regular get job, these parameters are required:
remote-file
The pathname of the file to download from the remote server.
local-file
The pathname to use on the local server for the downloaded
file.
For a regular put job, these parameters are required:
local-file
The pathname of the file to upload to the remote server.
remote-file
The pathname to use on the remote server for the uploaded file.
For a recursive get job, these parameters are required:
remote-file
The pathname of the file or directory to download from the
remote server.
local-dir
The directory pathname to use on the local server to contain
the downloaded items.
For a recursive put job, these parameters are required:
local-file
The pathname of the file or directory to upload to the remote
server.
remote-dir
The directory pathname to use on the remote server to contain
the uploaded items.
The rest of the parameters are optional. The spooler will attempt to
use reasonable defaults for these parameters if necessary.
user The username to use to login to the remote server. Defaults to
‘‘anonymous’’ for guest access.
pass The password to use in conjunction with the username to login
to the remote server.
acct The account to use in conjunction with the username to login to
the remote server. The need to specify this parameter is
extremely rare.
port The port number to use in conjunction with the remote hostname
to connect to the remote server. Defaults to the standard FTP
port number, 21.
host-ip The IP address to use in conjunction with the remote hostname
to connect to the remote server. This parameter can be used in
place of the hostname parameter, but one or the other must be
used. This parameter is commonly included along with the
hostname parameter as supplemental information.
xtype The transfer type to use. Defaults to binary transfer type
(TYPE I). Valid values are I for binary, A for ASCII text.
passive Whether to use FTP passive data connections (PASV) or FTP
active data connections (PORT). Valid values are 0 for active,
1 for passive, or 2 to try passive, then fallback to active.
The default is 2.
recursive
This can be used to transfer entire directory trees. By
default, only a single file is transferred. Valid values are
yes or no.
delete This can be used to delete the source file on the source
machine after successfully transferring the file to the
destination machine. By default, source files are not deleted.
Valid values are yes or no.
job-name
This isn’t used by the program, but can be used by an entity
which is automatically generating job files. As an example,
when using the -bbb flag with ncftpput, it creates a job file
on stdout with a job-name parameter so you can easily copy the
file to the job queue directory with the suggested job name as
the job file name.
pre-ftp-command
post-ftp-command
These parameters correspond to the -W, and -Y options of
ncftpget and ncftpput. It is important to note that these
refer to RFC959 File Transfer Protocol commands and not shell
commands, nor commands used from within /usr/bin/ftp or ncftp.
pre-shell-command
post-shell-command
These parameters provide hooks so you can run a custom program
when an item is processed by the spooler. Valid values are
pathnames to scripts or executable programs. Note that the
value must not contain any command-line arguments -- if you
want to do that, create a shell script and have it run your
program with the command-line arguments it requires.
Generally speaking, post-shell-command is much more useful than
pre-shell-command since if you need to use these options you’re more
likely to want to do something after the FTP transfer has completed
rather than before. For example, you might want to run a shell script
which pages an administrator to notify her that her 37 gigabyte file
download has completed.
When your custom program is run, it receives on standard input the
contents of the job file (i.e. several lines of variable=value key-
pairs), as well as additional data the spooler may provide, such as a
result key-pair with a textual description of the job’s completion
status.
post-shell-command update a log file named /var/log/ncftp_spooler.
#!/usr/bin/perl -w
my ($line);
my (%params) = ();
while (defined($line = <STDIN>)) {
$params{$1} = $2
if ($line =~ /^([^=\#\s]+)=(.*)/);
}
if ((defined($params{"result"})) &&
($params{"result"} =~ /^Succeeded/))
{
open(LOG, ">> /var/log/ncftp_spooler.log")
or exit(1);
print LOG "DOWNLOAD" if ($params{"op"} eq "get");
print LOG "UPLOAD" if ($params{"op"} eq "put");
print LOG " ", $params{"local-file"}, "\n";
close(LOG);
}
DIAGNOSTICS
The log file should be examined to determine if any ncftpspooler
processes are actively working on jobs. The log contains copious
amounts of useful information, including the entire FTP control
connection conversation between the FTP client and server.
BUGS
The recursive option may not be reliable since ncftpspooler depends on
functionality which may or may not be present in the remote server
software. Additionally, even if the functionality is available,
ncftpspooler may need to use heuristics which cannot be considered 100%
accurate. Therefore it is best to create individual jobs for each file
in the directory tree, rather than a single recursive directory job.
For resumption of downloads to work, the remote server must support the
FTP SIZE and MDTM primitives. Most modern FTP server software can do
this, but there are still a number of bare-bones ftpd implementations
which do not. In these cases, ncftpspooler will re-download the file
in entirety each time until the download succeeds.
The program needs to be improved to detect jobs that have no chance of
ever completing successfully. There are still a number of cases where
jobs can get spooled but get retried over and over again until a
vigilant sysadmin manually removes the jobs.
The spool files may contain usernames and passwords stored in
cleartext. These files should not be readable by any user except the
user running the program!
AUTHOR
Mike Gleason, NcFTP Software (http://www.ncftp.com).
SEE ALSO
ncftpbatch(1), ncftp(1), ncftpput(1), ncftpget(1), uucp(1).