I dug up some docs and found the format is Tape label Fileset label First backupset Filesetlabel second backup set
where each section/filemark begins a tar archive. Tape label is a file in Tar format that allows the program to identify the tape. Data following tape label are fileset/backup pairs that contain the data archived from a backup. If I cat or less the archive file created by the dd you suggested I get some test info about the fileset in ascii (see below). If I less the archive I get this info plus what looks like binary data. However, a tar -tvf of the tape or archive file from dd just gives me the /tmp/tapexx_lbl. The block size is 240 for the backup program and the tape is set to a blocksize of 0 for the SCSI tape. mt show this. SCSI 2 tape drive: File number=34, block number=0, partition=0. Tape block size 0 bytes. Density code 0x28 (Exabyte Mammoth-2). Soft error count since last status=0 General status bits on (81010000): EOF ONLINE IM_REP_EN This was designed to be retrieved by standard tar utilties but I guess I'm not using tar right <G>. You mentioned changing the dd block size - any suggestions? /tmp/fs_95.lbl01006600000000000000000000003712101477727110007015 [Fileset Label] FS_BACKUP_TYPE=4 FS_START_DIR=/ FS_X_COMMAND=-fcbFVSa Library1 240 /usr/bp/lists.dir/sub95.inc -zSTATION=gandalf -zWHERE=/ -zVERIFY=2 FS_BLOCKSIZE=240 FS_NODE_NAME=gandalf On Tuesday February 7 2006 10:20, Richard Fish wrote: > On 2/5/06, Brett I. Holcomb <[EMAIL PROTECTED]> wrote: > > Okay, I think I figured out what they are doing. They have a bunch of > > files for the labels. If I move forward using asf n where n is a number > > from 1-n I can walk through the label files. They take two files/label > > file so I go from 1 to 3 to 5 .... > > > > How do I get to this file to untar it? What I have is this when I do tar > > Thanks. > > Sorry for the slow response on this. > > It sounds like you don't really know the exact contents of the tapes, > so I think you should do something like: > > # dd if=/dev/tape0n of=archive1 bs=10k > # dd if=/dev/tape0n of=archive2 bs=10k > ... > # dd if=/dev/tape0n of=archiveN bs=10k > > This should give you a dump of all of the data on the tape, and then > you can analyze it in more detail. You might have to fiddle with the > bs= value above though. > > For some background info, tape devices generally write file marks > between archives. So as long as you are using the no-rewind tape > device and reading the full archive, you can usually just read them > one after the other. The mt fsf command is mostly useful for skipping > over archives. > > However, tape devices are not very consistent. Sometimes if you read > just part of an archive and close it, the tape will automatically move > to the next file mark. Other devices will require an mt fsf command > to get to the next file mark. > > The asf command sometimes works, and sometimes doesn't. rewind and > fsf is the safest method. > > -Richard > > > On Sunday February 5 2006 23:36, Richard Fish wrote: > > > On 2/5/06, Brett I. Holcomb <[EMAIL PROTECTED]> wrote: > > > > I have a scsi tape library and a backup program that creates datasets > > > > of tar files on the tapes. I gather each dataset is a tar file. I > > > > would like to be able to access each of these tar files. At this > > > > point I can tar -tvf /dev/tape0 and see the file that contains the > > > > tape label. But I can't get beyond that. I've tried skipping to the > > > > next file, records, set mark using mt with no luck. > > > > > > mt is the correct command, but you need to make sure you are using a > > > no-rewind tape device (ntape or nst0). Otherwise you will end up > > > seeking to the next file, closing the file descriptor, which causes > > > the driver to rewind the tape. > > > > > > -Richard > > > > -- > > > > Brett I. Holcomb > > -- > > gentoo-user@gentoo.org mailing list -- Brett I. Holcomb -- gentoo-user@gentoo.org mailing list
