Amanda-Users

Re: Backups that don't fit on the tape....

2004-02-12 13:14:42
Subject: Re: Backups that don't fit on the tape....
From: Joshua Baker-LePain <jlb17 AT duke DOT edu>
To: Mike Brodbelt <m.brodbelt AT acu.ac DOT uk>
Date: Thu, 12 Feb 2004 13:12:30 -0500 (EST)
On Thu, 12 Feb 2004 at 6:02pm, Mike Brodbelt wrote

> Yes - rather less:-
> 
> Tape Time (hrs:min)        0:35       0:35       0:00
> Tape Size (meg)          8461.0     8461.0        0.0
> Tape Used (%)              24.2       24.2        0.0
> Filesystems Taped            15         15          0
> Avg Tp Write Rate (k/s)  4179.2     4179.2        --
> 
> > Your later posting showed the amount actually taped when the error occured:
> > 
> > == taper: tape ACU-STD-2 kb 23940992 fm 16 writing file: Input/output error
> > 
> > Only 23.94 GB and 16 filemarks when the error occured.  Hardly 30.5 GB.
> 
> 23.94 Gb is the size of the compressed dump image for the single large
> filesystem that failed to tape. 16 filesystems dumped, but only 15 taped
> successfully.

23.94 GB is *also* the amount of data that actually went to tape before 
amanda got the I/O error.  You indicated that these "should" be 35GB.  The 
fact that amanda only got 23 out to tape indicated either that hardware 
compression is on, you're using smaller tapes, or there's something wrong 
with your hardware.  For the record, this is the message I get when 
hitting EOT:

  taper: tape RAIDsSet2-10 kb 98708224 fm 22 writing file: No space left 
on device

which is obviously *not* the error you got.

> The previous poster suggested that my hardware was faulty due to the I/O
> error. AFAIK, it's working OK, which is why I was looking for
> clarification on that. ISTM that the error above, combined with the "out
> of tape" line at the top of the report make it reasonable to assume that
> the I/O error was just amanda hitting EOT when she thought there should
> still be tape left. I suspect this is due to my having lifted the
> tapetype definition from the FAQ-o-matic originally, and the size in the
> tapetype being over-optimistic. Is that a reasonable interpretation, or
> should I be pulling the machine to bits looking for dodgy connectors?

Amanda only uses tapelength to plan.  After it forms its plan it just 
dumps everything it has planned to to the tape.  Sometimes this gets 
bigger between estimate and dump time, and thus you hit EOT.  Or sometimes 
the tapelength is bigger than the actual tape size, and you hit EOT.  But 
you don't seem to be hitting EOT.

Check your hardware thoroughly.  Clean the drive.  Check the compression 
status definitively using 'tapeinfo', part of the mtx software package.

And it's certainly looking like it's time to get a goat.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University