Re: amdump & log files from 2.6.1 run yesterda
2009-02-20 16:06:40
FYI, I've also been seeing this error/issue with 2.6.1. My setup is
simpler (single manual tape drive), and it seems to be occurring when
the dumps fill a tape. Looks like the wrong error is making it's way
out of the taper for some reason. It also looks like dumps (or
possible dumps which the taper was trying to tape as partials) are not
being removed from the holding disk, so it keeps trying and building
them up on the holding disk. My taper debugs look similar to Stan's,
but I can post them if you're interested. I cleared out an overlarge
dump that had been tried as a partial tape run twice, and used amflush
to clear out a bunch more (which threw the same error when it filled
the tape, btw), and last night it ran clean again (and only filled
about 60% of a tape!). I'll be keeping an eye on it, and will let you
know if it happens again.
-Darrell
On Feb 20, 2009, at 1:44 PM, stan wrote:
On Fri, Feb 20, 2009 at 12:32:31PM -0500, Dustin J. Mitchell wrote:
On Fri, Feb 20, 2009 at 11:13 AM, Debbie O'Connell
<docrtp AT yahoo DOT com> wrote:
We think these are the files you wanted us to send to you. amdump.
1 (vs
amdump) had a timestamp from an earlier run yesterday, amdump is
from the
3:30pm testing we did that corresponds to the taper debug file.
btw, we
restored 2.5.2p1 to the server for last night's backup & it
successfully ran
using the tape in question - the vtape, logical & physical tape,
same drive,
etc. that fails under 2.6.1 was used for the nightly backup. It
found the
tape labeled as "DailyDump04" -> vtape5, slot5 (uses single tape
drive
/dev/nst0) & wrote everything successfully. Thanks, Debbie
Hmm -- I can't find anything in those logfiles that would indicate
why
the device is failing. The problem is really very specific, and does
not involve changers, but rather the RAIT, VFS, and Tape device
drivers.
IIRC you built from source. If I send a patch that adds additional
debugging information, could you try it?
Absolutely.
For reasons that _might_ be hsitorical we _always_ build Amanda from
source
on _every_ platform, At least historically man, many things that aree
(IMHO) site specifc, are/were confiure time options.
--
One of the main causes of the fall of the roman empire was that,
lacking
zero, they had no way to indicate successful termination of their C
programs.
|
|
|