It neither works, then you must use dd!

For each file on tape
  position the tape on the file
  dd if=/dev/nst0 of=file.[number] bs=32k skip=1
concatenate all file.[number] files.

'amadmin [config] find hostname /fileset' list all files on tape you need.

You can also try to upgrade the server to 2.6.1p1.

Jean-Louis

Franck GANACHAUD wrote:

Used amfetchdump and after 3hours of restore 10M files, it just throwed a

amfetchdump: 3006: restoring split dumpfile: date 20090424 host hostname disk /fileset part 2992/UNKNOWN lev 0 comp .gz program /bin/tar

amfetchdump: missing file header block
could not fsf /dev/nst0: Input/output error
amfetchdump: could not fsf /dev/nst0: Input/output error

What does mean? I checked the library and it's still on first tape.
What can I do to manually get what's on various tapes?

Btw, I changed the chunksize to 1Gb in amanda.conf as Dustin suggested.

Jean-Louis Martineau a écrit :

It is easier to restore a splitted dump with amfetchdump. It will write a 250Go file.

There is no way to automatically rebuild index from tape.

Jean-Louis

Franck GANACHAUD wrote:

Hello,

Server and client are 2.5.1p1 on debian etch

I've been trying to use amrestore but the backup is around 250Go splitted over 5 tapes and I'm not successful so far. amrestore is restoring the fileset using 10M size files. Problem is amrestore doesn't close 10M files it already restored (can see it using lsof) and after some time, I hit the system limit for open files. I rose the system limit *4 but I'm not sure it will be enough as 250Go/10Mo is a pretty high number.

Is this situation normal ?

Btw, is there a way to rebuild indexes from tapes?

Franck.