On Mon, Nov 30, 2015 at 04:52:22PM +0000, David Simpson wrote: > Sorry I'm back so soon with more Qs - I guess getting stuck into the > restores is giving me a lot to think about! > > How can one find out the number of filemarks on the tape? (assuming > worst case no amanda info and best case) > -Do I need to know this? (see below) > > ----- > > Layout: > > How is Amanda writing to the tape? > -I have part_size 200G > -Blocksize 1024k > -TAR > > Does this mean I have this layout(?): > > | 1024k | Start of DLE or DLE up to 200GB in 1024k blocks | > 1024k | DLE or continuation of DLE, 200GB in 1024k blocks | > ---------------------------------------------------------------------------------------------------------------------------------------------------------------------> > | HEADER | AMANDA DUMP (UNDERSTOOD BY TAR) | HEADER > | AMANDA DUMP (UNDERSTOOD BY TAR ) | >
I think there is a small misunderstanding. amanda writes a set of "tape files" (TFs). There may be a filemark after each TF (see later). Each TF is written is blocksize blocks (1024K in your example) The first TF on the tape is the amanda "Tape Header". The Tape Header will be one block but probably only a little data at the start then nulls. Each remaining TF will be either a complete or partial dump. If splitting is allowed, the largest a TF will be is part_size (200G in your example). Each of these TF (partial or complete) will start with an amanda "File Header". Like the Tape Header, it will be 1 block in size with only a little data at the start. Note, the File Header is part of the TF, not separate as in your drawing. Thus, it must be stripped before the amdump data is fed to the recovery program (eg. tar or dump). Filemarks may not be a consideration. Some tape formats (eg. DAT) actually wrote, at the end of each TF, a small section of tape with special code saying this is the EOF. Thus when fast forwarding it could count "filemarks" to know when it reached the file you requested. These filemarks consumed tape that could have held user data and sometimes were sufficiently large that they were considered by amanda in its planning. Eg. how many small incremental dumps will I make with each adding its filemark? I think all recent tape formats keep separate tape tracks for metadata such as block size, block compression status, and EOF marks. On these formats the filemark does not consume space on the user data tracks and thus may not be important to amanda planning. Even if "tapetype" does report a filemark size > 0, it is typically so small with respect to the tape capacity to be negligible. Jon -- Jon H. LaBadie j...@jgcomp.com 11226 South Shore Rd. (703) 787-0688 (H) Reston, VA 20190 (703) 935-6720 (C)
