ras - redundant archive system

ras [-r] [-b|-v|-s] [-HF] [-f segfile] ... [-n segcount] [sumfile] ...

ras -c [-b|-v|-s] [-HF] [-f segfile] ... [-n segcount] [-z blocksize] [-o offset] [sumfile[:sumspec]] ...

ras adds m extra files to a set of n files, such that the contents of the n original files can be regenerated from any n of the n+m original files and extra files. It works for any values of n and m so long as n+m<=256. The extra files will normally be 6 bytes larger than the largest of the original files, but they can be made the same size as the largest segfile.

The original files are termed "segment files" or "segfiles", and n (the number of segfiles that existed before any were lost) is termed the "segcount". The extra files created by ras are termed "sumfiles", and m (the number of sumfiles created) is termed the "sumcount". Each sumfile has an ID, an integer from 0 to 255-segcount. There's not much point having 2 sumfiles with the same ID, since they would contain the same information ... you would essentially have 2 copies of the same sumfile.

-r,--restore

Restore mode, restore any missing segfiles using the sumfiles specified on the command line. If there are surplus sumfiles (i.e. more than the number of missing segfiles) then only the first few will be used unless some of the first few turn out have the same IDs. This is the default mode of operation.

-c,--create-sums

Create mode, create the sumfiles specified on the command line. The sumfile arguments may be simply filenames, in which case the first one will be created with ID 0, the second one will be created with ID 1, and so on. The sumfile arguments can optionally include a sumspec (a number specifying the ID to use) separated from the sumfile filename by a ":". The sumfile filename, the colon and the sumspec should be a single argument with no spaces.

-b,--brief

Output only a 1-line summary of what's going on to standard output. This is the default level of verbosity.

-v,--verbose

Output some more detailed information to standard output.

-s,--silent

Output nothing to standard output.

-f,--filename segfile

Use repeatedly to set the names to be used for segfiles, for example "-fseg1 -fseg2 -fseg3" if there are 3 segfiles named seg1, seg2 and seg3. By default ras uses "xaa", "xab", "xac", ..., "xay", "xaz", "xba", "xbb", ... as the names of the segfiles. This convention was chosen because it's the same as that used by the GNU version of split(1)

-n,--segcount segcount

This option allows the segcount to be specified on the command line. This option is not normally needed since ras can determine the segcount by testing for the existence of segfiles when creating sumfiles, and by looking at the header of one of the sumfiles when restoring segfiles.

-o,--offset offset

When creating sumfiles, this option causes ras to start from offset instead of 0 when assigning IDs to sumfiles. See EXAMPLES below. Ignored when in restore mode.

-H,--no-header

ras normally outputs a 5-byte header to each sumfile generated, consisting of the characters "SU8", a byte specifying segcount, and a byte specifying the ID of the sumfile. In create mode this option suppresses the writing of the header. In restore mode this option tells ras not to expect to find a header. If you use this option to create some sumfiles then you must also use it when using those sumfiles to restore segfiles. You can't mix sumfiles with headers and sumfiles without headers. If you use this option you must keep track of segcount and the sumfile IDs yourself. When restoring you must tell ras how many segments there are (using the -n option, above) and ensure that the names of the sumfiles reflect their ID values. For example, under the -H option, ras will assume a sumfile called "sum_45" has an ID of 45. Note that it takes the first block of digits that it finds in the sumfile name, so a sumfile called "s0er_34" will be treated as having an ID of 0, not 34. In summary, don't use this option unless you really need to get rid of the 5 byte header overhead.

-F,--no-footer

ras normally appends a 1-byte footer to each sumfile created. This option suppresses this footer in create mode, and tells ras not to expect to find it in restore mode. If you use this option to create some sumfiles then you must also use it when using those sumfiles to restore segfiles. You can't mix sumfiles with footers and sumfiles without footers. The function of the footer is to allow ras to handle segfiles of differing lengths correctly. If you don't have the footer, then if any segfile smaller than the largest is lost and restored, it will get padded with zeros to the length of the largest segfile. If all the segfiles are the same size, of course, this won't matter. If you use both the -H and the -F options, then the sumfiles will be exactly the same size as the largest of the segfiles.

-z, --block-size blocksize

This option sets the size (in bytes) of the blocks in which ras processes data. This option has no effect on the data that's output, it's provided for performance tweaking only. The default block size is 1024 bytes.

-V,--version

Print version information and exit

-h,--help

Print some help and exit

To generate sumfiles named sum0, sum1 and sum2 of segfiles named xaa, xab, xac and xad in the current directory, using sumfile IDs 0, 1 and 2:

$ ras -c sum0 sum1 sum2

As above, but there is also a file called xae in the current directory that we wish to ignore:

$ ras -c -n4 sum0 sum1 sum2

As above, but we wish sum0 to have an ID of 15, sum1 to have an ID of 1, and sum2 to have an ID of 2:

$ ras -c -n4 sum0:15 sum1 sum2

As above, but we wish sum0 to have ID 45, sum1 to have ID 46 and sum2 to have ID 47:

$ ras -cn4 -o45 sum0 sum1 sum2

As above, but the segfiles are called blk1, blk2, blk3 and blk4.

$ ras -cn4 -o45 -fblk1 -fblk2 -fblk3 -fblk4 sum0 sum1 sum2

Note that we still need the -n4 in that last example, since we haven't specified the name of the 5th segfile, so it stays at the default of xae, which exists.

We want to restore 2 missing segfiles from a set of 5 named xaa, xab, xac, xad and xae using whichever of the sumfiles named s0, s1, s2, s3 and s4 are available:

$ ras s?

As above, but there may be no sumfiles, in which case we wish ras to exit successfully if all the segfiles are there and to fail if any are missing.

$ ras -n5 s?

Now for a slightly longer example: suppose we have a large file that we wish to transport between two machines using 1.44Mb floppy discs. I'm assuming that we've already compressed it, added a checksum and split it into 1400000-byte chunks. The checksum is advisable because it's possible that there's a bug in ras that will silently corrupt the data. So, for example, we have three files of size 1400000 bytes (called xaa, xab and xac) and a file of size 800000 bytes called xad. We could put these on 4 floppies, but instead we choose to have some redundancy, and use 6 floppies such that any 4 of the 6 will suffice. We create 2 sumfiles...

$ ras -c sum0 sum1

Now we have 2 extra files, sum0 and sum1, both of size 1400006 bytes. We add individual checksums to each of our 6 files and copy them to 6 floppies. Suppose that when we get to the destination machine, we find that the floppies holding xab and xac are corrupt. We can use ras to regenerate xab and xac from xaa, xad, sum0 and sum1.

$ ras sum?

And finally we concatenate xa[abcd], check the checksum and so on. For a more concrete example see the shell scripts rassplit and rasmerge, (distributed with ras) for splitting a large file across multiple floppies with redundancy.

rassplit(1), rasmerge(1)

Nick Cleaton <nick@cleaton.net>