Bacula-users

Re: [Bacula-users] (it was Remote storage) LTO3 is not working, needing urgent help

2008-12-26 18:24:24
Subject: Re: [Bacula-users] (it was Remote storage) LTO3 is not working, needing urgent help
From: John Jorgensen <jorgnsn AT lcd.uregina DOT ca>
To: "John Drescher" <drescherjm AT gmail DOT com>
Date: Fri, 26 Dec 2008 17:20:56 -0600
>>>>> "drescherjm" == John Drescher <drescherjm AT gmail DOT com> writes:

    >> Now, I doing :
    >> 
    >> mt -f /dev/nst0 erase
    >> 
    >> But I issue this command at 8:07 AM GMT, and now is 9:44 AM GMT, is it
    >> common that it takes too much time?
    >> 
    drescherjm> That can take a very long time

    drescherjm> A better way is:

    drescherjm> mt -f /dev/nst0 rewind
    drescherjm> mt -f /dev/nst0 weof
    drescherjm> mt -f /dev/nst0 rewind

That's certainly a much faster method, and so better in the case
where you're just trying to rewrite the tape yourself.

I can think of one rare circumstance, though, where "mt erase" is
probably better than "mt weof" (I say "probably" because I'm
about to spread some of my own speculation, based on external
observations made while retiring some old DLT tapes, rather than
on an insider's knowledge of how tape drives work internally).

I assume that "mt erase" takes so long because it actually is
clearing the entire length of the tape, rather than just putting
an end-of-data marker at the beginning of the tape, while
leaving the original data undisturbed on the far side of that eod
marker.  (I observed that when I issued 

  mt -f /dev/nst0 erase

the tape drive's "in use" LED blinked for about 2 hours,
the same amount of time it blinked if I ran 

  dd if=/dev/zero of=/dev/something bs=126b

where "/dev/something" was a special device name to open the
drive in uncompressed mode).  

So, in principle, the data after the end-of-data mark written by
"mt weof" might still be readable with somebody with the
wherewithal to write a custom device driver to force the tape
head past the eod mark, and almost certainly would still be
readable to someone with the ability to disassemble the tape
cartridge and build their own custom raw tape reader.

I doubt most of us have data that anybody would want to steal
badly enough to build custom hardware for the job, but there
probably are quite a few shops whose official procedures require
that erased data actually be overwritten before media is
discarded, unless the tapes are destroyed.

Incidentally, there was another bit of circumstantial evidence
supporting this hypothesized distinction between "mt erase" and
"mt weof".  Running "mt eod" on a tape which had just been
submitted to "mt erase" caused our DLT drive to sit thinking for
15 minutes or so before it would accept any other commands, while
"mt eod" after "mt weof" returned immediately.  If I remember
correctly, a posting on backupcentral (which I can't find again
now) speculated that the erase command overwrote a directory
between tape files and tape blocks, hence leaving the drive lost
when it initially reloaded the tape.

-- 
John Jorgensen  LCD System Administrator  jorgnsn AT lcd.uregina DOT ca
                                          306.337.2344

------------------------------------------------------------------------------
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>