User talk:Freelsjd

tarcat

Hi there -- I just reverted the tarcat change on the front page because I don't see how it's related to Amanda. When Amanda splits dumps into multiple on-tape files, it does so by breaking the tarfile without regard to its contents (that is, without tar's knowing about it to add a new header) and adding an Amanda header. Furthermore, even if tarcat could reassemble these files, Amanda already has a utility to take care of this -- amfetchdump can reassemble split dumps into a single tarfile as necessary.

Please don't feel that my revert is final or irreversible. Rather, let's discuss the relevance of the utility first, and then we can look at ways to include it in the Amanda distribution, in the manpages, or elsewhere in the wiki. For maximal community involvement, either of the amanda-users or amanda-hackers lists would be best, but please also feel free to carry on the discussion here. --Dustin 18:08, 5 July 2007 (PDT)

Thanks for the feedback Dustin. Indeed, I did try to use amfetchdump as my first attempt at recovery. The problem is that amfetchdump requires the index/info files from AMANDA in order to do it's job. These files were on the failed drive, so I could not use amfetchdump. I was able to perform a new install on a new hard drive and install AMANDA. Then I could use amrestore to get the files. I could look at my save e-mail messages from AMANDA and determine which tapes to use to run amrestore. What I then had was a set of tar files that must be combined in order to create the contiguous tar file needed to restore my root et al. This is where tarcat came in handy. I also looked at tarcat and this is what I actually used. I think that once amrestore does it's work, the AMANDA header is no longer present and you are left with a pure tar file (or set of tar files). The tar files are also split into multi-volumes that cannot be combined any other way that I could determine (what tarcat did).

I disagree and think tarcat is very valuable tool to have in the AMANDA arsenal. I am just trying to help others also have it. -- Jim, 09:43, 6 July 2007 (EDT)

Ah, I see. As I mentioned above, Amanda does not use multi-volume support. A multi-volume tarfile contains the datastream from a single tarfile, but each constituent "chunk" of the multi-volume is prefixed by a Tar header. Examines this header and strips it off. Instead, Amanda takes one very long tarfile and chops it up arbitrarily, adding an Amanda header to each one. As you mentioned, amrestore strips those Amanda headers off, leaving you with a set of files which, concatenated together, are a single tarfile. That is, the first file has a Tar header, and the rest have no header at all.

Tarcat tries to recognize 'tar' headers and, when it doesn't find one, it just acts like 'cat'. So in the case of joining the results of amrestore, tarcat chunk1 chunk2 chunk3 is exactly equivalent to cat chunk1 chunk2 chunk3, although in some cases would mis-identify raw data as a Tar header (in which case it would produce a corrupt tarfile by stripping blocks from the beginning of the file).

We should make a How Tos page for recovering from a failure of the drive containing the Amanda indices.

--Dustin 08:41, 6 July 2007 (PDT)

Thanks Dustin. I agree with you now. I wondered if cat would do the trick, but then found tarcat.

Well, thanks for contributing! Please feel free to join the mailing lists or contribute documentation of your recovery process to the wiki. --Dustin 09:27, 6 July 2007 (PDT)