On Thursday 06 July 2006 14:41, Jon LaBadie wrote:
> On Thu, Jul 06, 2006 at 10:49:22PM +0200, Peter Kunst wrote:
> > Hi list,
> >
> > ever seen something like this in amstatus output ?
> >
> >  matrix:/var/lib/backuppc 0 planner: [hmm, disk was stranded on waitq]
> >
> > This is a new debian box, which (guess it :) runs BackupPC, and was
> > taken into amanda's disklist today. This DLE should have around ~60GB.
> >
> > On this client, there is a "tar [foobar] --file /dev/null [barfoo]"
> > still running now, and i guess amanda runs into etimeout. However, never
> > seen this message since running amanda (~3 years now). Not an urgent
> > issue, since i'm just moving over amanda from an old amanda-server to a
> > new one with another robot attached.
>
> A tar run to output file "--file /dev/null"
> is part of the estimate phase.
>
> Tar senses the /dev/null output and only looks
> at directory blocks and inodes.  Doesn't read
> the data blocks for other files.
>
> Never saw the "stranded" message.
> Wonder which tropical island it is on :)

I've encountered the "stranded" message.  In my case there were several DLEs 
for that host.  One of them was taking forever to dump, which was blocking 
the dumps of the other DLEs.

As it turned out, the LONG dump time in my case was due to /var/log/lastlog (a 
sparse file with apparent size of 1.2TB on a 64bit machine).  tar was unable 
to deal with it.  The only solution I found was to explicitely exclude it.

-- 
--------------------------------------------------------------
Jeffrey Anderson                        | jdanderson AT lbl DOT gov
Lawrence Berkeley National Laboratory   | 
Office: 50A-5104E                       | Mailstop 50A-5101
Phone: 510 486-4208                     | Fax: 510 486-6808