Each computing environment at the ATCA handles exabytes and cdroms somewhat differently. Apart from the obvious differences between the VMS and UNIX machines, the three flavours of UNIX (Digital UNIX, SunOS 5.x and SunOS 4.x) differ in the details.

ATCA data is archived in RPFITS files on cdroms or exabyte tapes. The format for the latter is an ANSI tape, which is structured with a header, the data, and a trailer for each RPFITS file written on it. A useful utility to access these tapes is ansitape (see also ansiread below). This can be used to get directory listings of the tape as well as to read the contents onto disk, and even to write an ANSI tape if desired. Note that if you read the data to disk with this type of utility, you must still read the file into some package such as Miriad or AIPS with their filler programs (called atlod) for further work.
 

Usage

There are a number of devices available on a variety of computers:

• Vladimir (the Archiving Workstation ) is used for archiving raw RPFITS data direct from the correlator (exabyte and cdrom).
• Kaputar (the compute server) is used for reading/writing exabytes and reading cdroms.
• Many of the Sun workstations and PCs can read cdroms, and
• Aine is an off-line cdrom writer.
• Leon (the off-line VMS machine) also has a exabyte reader, but this is rarely of general interest.

The off-line machine used to write CDs is aine. There is a separate document describing using the cdrom writer.
 
 

Mounting and Reading CD-ROMs on the UNIX machines

Although the details of mounting and unmounting differ, once mounted a cdrom appears as a normal file system to the user (although it is read-only). Cdroms are infinitely more convenient to deal with than exabytes - their use is strongly encouraged.

Here we summarise how to mount and unmount a cdrom:

• Kaputar: There are three drives: drive 0, 1 and 2. These are in the observers area near the printers. Mount with the command

% cdmount x

where x is 0 or 1. The mounted cdrom appears at /cdrom0, /cdrom1 or /cdrom2. To unmount, use

% cdumount x

where again x is 0 or 1.
• Achilles, Medea and Ambrosia: Their drives are located in the CPU box. Once placed in the reader, the cdrom will be mounted automatically (it takes about a minute for this to happen, so be patient). The mounted cdrom appears at /cdrom/xxxx, where xxxx is the cdrom volume name. To unmount/eject, use

% eject cdrom

• Vladimir: The cdrom drive is located on the right-hand side of the workstation. The there is a single command which both mounts and unmounts/ejects the cdrom, namely

% cdswap

The mounted cdrom appears as /cdrom/xxxx, where again xxxx is the volume name of the cdrom.

Because cdrom files are read-only, data copied from cdroms will also usually be read-only. You can change this with the chmod command.
 
 

The Exabyte Device on leon

Leon's exabyte drive is located in the rack on the left as you enter the Computer Room (immediately above leon itself). VMS read and write ANSI tapes by default. You can mount an initialised tape with

$ allocate mka600:
$ mount/override=ident mka600:

$ copy mka600:96-07-29_0213.C537 []

$ initialize mka600: xxxx

There is no software mounting of tape devices in Unix so you access the drive directly. The form of the device names differs between the UNIX flavours. There are two tape interfaces of interest here - the so-called `non-rewinding raw' and `raw' interfaces. Different programs require different interfaces. The following table summarises the devices available at the ATCA.
 

Other variations on these names are possible. See the man page on mtio for more information (although this is at best cryptic).

Kaputar has two exabyte drives, which are in the observers area near the printers. Each is labelled with its ``non-rewinding raw'' name. Vladimir's drive is adjacent to the workstation, it can handle 2 and 5GB exabytes. See the documentation on Exporting ATCA data to find out how to get your data onto the Vladimir exabyte.
 
 

Reading and Writing Exabytes on the UNIX machines

There are basically two sorts of tape format that are used in the UNIX world - tar and ANSI format. Tar is the usual format for exchange between different UNIX environments. However, ANSI tapes are also sometimes used. In particular, exabytes written by the ATCA archiving software are in ANSI format. Utilities to read and write exabytes are as follows:

• atlod - AIPS and Miriad atlod tasks can read and translate ATCA ANSI RPFITS exabytes directly. Copious information on these exist in the ATNF AIPS and Miriad manuals. AIPS requires a tape number (always 1 at the ATCA), whereas Miriad requires a non-rewinding raw device name. These are available on kaputar only.
• ansitape - This reads and writes ANSI tapes. It is available on Vladimir (archiving workstation) and Kaputar. This is the preferred way to read/write ANSI tapes on UNIX machines. Note that ansitape requires a `raw' (not non-rewinding raw) tape interface.


ansitape's command line inputs are somewhat unusual. For example, the following command gives a directory listing of the tape on Vladimir:

% ansitape t mt=/dev/rst0

Filenames on ANSI tapes are case-insensitive -- ansitape treats them as if they are lower case, so, e.g., to extract all the files on the Kaputar tape whose names end with .C234, use the following command

% ansitape xv mt=/dev/rmt1a \*.c234

Note that we need to 'escape' the "*", to keep the unix shell from interpreting it before ansitape gets to it. See ansitape's man page for more detailed information.

• ansiread - This reads ANSI tapes and is available on kaputar only. It is simpler-minded (and simpler to use) than ansitape. However note that it truncates filenames to 17 characters (ATCA RPFITS files usually have 18-character names). Also unlike ansitape, you must use a non-rewinding raw device interface. For example,


% ansiread -f /dev/nrmt1a -s 2 -r 4

skips 2 files (note that ansitape does not have this file skipping ability) and then extracts 4 files to disk. Their names will be constructed from the ANSI headers. See its man page for more information.

• tar and gtar - These are the traditional UNIX utilities to read and write tapes, and have similar functionality (gtar is the GNU version of tar). Apart from vladimir missing gtar, both programs are available on the different UNIX systems at the ATCA.


Note that the tar format written by kaputar cannot be read by SunOS 4.x versions of tar. For safety's sake, it is suggested that gtar be used on kaputar.

tar can recursively copy full directory trees and it is quite fast. One thing to be avoided is writing files with absolute paths. If you do this, that full path must exists when you extract files from the archive somewhere else. Always try and use relative paths. For example, to archive all the files in the home directory of user pwild onto an Exabyte tape we would issue the commands

% cd ~pwild
% tar -cvf /dev/nrmt1a .

which instructs tar to act verbosely (switch v) whilst it creates (switch c) an archive on the device /dev/nrmt1a (switch f), acting recursively (default) from the current working directory (the .). DO NOT use a command with an absolute path such as

% tar -cvf /dev/nrmt1a /narusers/pwild

To extract files from a tar archive, the inverse operation is

% tar -xvf /dev/nrmt1a

where the x means extract. This will create any subdirectories as needed below the working directory.

You can list the files on the archive with the command

% tar -tvf /dev/nrmt1a

You can select a subset of the files on the archive by naming the files; you must name them with the full relative path that the above listing command gives. For example, let us say that the above example which lists the files in the archive gives the following output as part of its listing:

rw-r--r--503/100  29812 Aug 17 19:24 1959 ./ideas/jsquared.txt  
rw-rw-r--503/100   7088 Jan 22 18:52 1961 ./ideas/heliograph.txt  
rw-------503/100   7196 Feb 12 18:52 1982 ./ideas/closeit.txt  

This indicates that there is a subdirectory called ideas beneath the directory from which we made the archive (we were in the pwild home directory, and accessed it by calling it . when invoking tar). In this directory are three files. If we want to extract the one called heliograph.txt then we would issue the command

% tar -xvf /dev/nrmt1a ./ideas/heliograph.txt

which would create the ideas subdirectory below our current working directory, and then fill it with the desired file.

Finally, tar archives are a convenient way to move files around if you have a lot of directories. The archive may be a disk file rather than a tape record. For example, the following commands create a compressed tar archive of a home directory tree ready for transport (perhaps by ftp) to another computer for reinstallation. We write the archive in some other directory of course to avoid recursive loops.

% cd ~pwild
% tar -cvf /DATA/HYDRUS_1/other/pwild/home.tar .
% cd /DATA/HYDRUS_1/miriad/pwild
% compress home.tar

which would leave behind a file called home.tar.Z ready to be expanded with uncompress somewhere else.

For more information, see the man page on tar, or use

% gtar --help

for a summary of gtar options.

The non-rewinding raw device interface should be used with tar programs.

• mt - This is a useful utility for manipulating the tape device. Here are some useful things you can do with it:


% mt -f /dev/nrmt1a eom
% mt -f /dev/nrmt1a fsf 6
% mt -f /dev/nrmt1a bsf 2
% mt -f /dev/nrmt1a rewind
% mt -f /dev/nrmt1a offline
% mt -f /dev/nrmt1a status

to wind the the tape to the end of the recorded media so that you can append some files, to skip forwards over 6 EOF marks, to skip backwards over 2 EOF marks, to rewind the tape, to rewind and eject the tape, and to get the tape unit to tell you where it thinks it is, respectively. See man mt for more details.

• dd - The basic Unix mechanism to extract information from a device is called dd. This can be used to dump data directly from tape to disk. The following example extracts a single fits file to disk and calls it junk.fits


% dd if=/dev/nrmt1a of=junk.fits ibs=2880

Note that you have to tell it the block size of the input record. This utility is very basic, and probably you won't need it. See man dd for more details.