>After completing the first dump cycle, I thought it was time to test Amanda, 
>to make sure that I could restore in case of emergency.  ...

Now what would possess you to do something silly like that?  :-)

>It's a good thing
>that I did. amrecover comes out with 'UNKNOWN file' whenever I try to 
>restore anything. I read the docs and tried to read the headers, and got this:
>
># mt rewind
># mt fsf NN
># dd if=/dev/rmt/0bn bs=32k count=1
>AMANDA: TAPESTART DATE 20010615 TAPE xx0
>
>1+0 records in
>1+0 records out
>
>Looks like the tape is hosed ...

This looks like the beginning of a tape.  Are you sure that fsf worked?

Two points:

  * Did you set the TAPE environment variable before doing the rewind
    and fsf commands?  If so, what to?  For instance, if it's set to
    a rewinding device name (e.g. /dev/rmt/0b), the fsf would advance
    out several files and then the tape would be rewound as the process
    exited.

  * Don't use the "BSD" semantic device names (the ones with 'b' in
    them).  Use (e.g.) /dev/rmt/0n.  Some portions of Amanda that deal
    with skipping across multiples files expect System V semantics.
    If you don't know what I'm talking about, count yourself lucky :-).

>the only clue I have is something about Variable Length I/O in syslog, if 
>anyone has a working st.conf for this drive in Sol2.6 then I can eliminate 
>that possibility.

I searched both the Tandberg and Sun sites and didn't find the appropriate
st.conf entry, but it shouldn't be too hard to come up with one.  Make a
copy of /kernel/drv/st.conf.  Then add this (which just duplicates the
Sun DLT7000 entry):

tape-config-list=
    "???",      "Tandberg DLT7000",     "DLT7k-data";

DLT7k-data =   1,0x38,0,0x1D639,4,0x82,0x83,0x84,0x85,2;

The tricky part is getting the vendor (first -- "???") inquiry text
string right.  It should be in your /var/adm/messages file right after
you did your first tape operation (which would cause the st driver to
load and scan the bus) and must exactly match the inquiry string returned
by the drive.

You can see that the right entry got picked in a couple of ways.
First, if it shows up as "Tandberg DLT7000" (the second text string)
in /var/adm/messages, it found the right match.  Another way is with
this test program that dumps a bunch of internal driver info:

  ftp://gandalf.cc.purdue.edu/pub/amanda/mtinfo21.c

>- John

John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]

Reply via email to