>I get an error when I copy over my 2.4.1 configs to 2.4.2 and try to
>run amcheck:
>
>   amcheck-server: could not get changer info: cannot open \
>      /opt/pkgs/amanda/etc/phg-weekly/changer.conf
>
>I don't reference a `changer.conf' file in any amanda.conf file.  I do
>reference chg-zd-mtx, though.  ...

Apparantly all (or at least chg-zd-mtx) changers now get their parameters
from "changer.conf".  This comment is in the chg-zd-mtx sources:

  # NOTE: all variables are now placed in a 'changerfile'.conf in your
  # amanda config directory, where 'changerfile' is what's set in
  # amanda.conf.  Ex. if amanda.conf has:
  #       changerfile="/etc/amanda/Dailyset1/CHANGER"
  # Then the variables file will be "/etc/amanda/Dailyset1/CHANGER.conf".

>I'm pretty sure it's ufsdump because my paper logs specify piping to
>ufsrestore.

OK.

>> Is any of this NFS mounted?  Even automounted from itself?
>
>Yes.  The partition that's having the problem is nfs mounted, however
>amanda is calling its daemon on the remote machine to back it up (ie,
>not thru nfs).  At least that's what I'm assuming is happening.

Let me make sure I've got this clear.  The host you gave to Amanda has
the "real" disk and is not getting it via NFS, right?  That would be the
correct way to do this, especially with ufsdump.

>> What do you mean by "/export/data/mail (linked to /var/mail)"?
>
>I just meant that /export/data/mail is symlinked to /var/mail, in case
>that could be a problem.  I'm specifying /export/data/mail as the
>partition to back up, so I don't think it is.

That's a problem.  You should give ufsdump (via Amanda) disk names (e.g.
/dev/rdsk/xxx) or mount points, never a subdirectory and certainly never
a symlink.  Ufsdump needs to convert whatever name you give it to a disk
because that's what it reads directly, and it does that via /etc/vfstab.
Since that name (symlink) is not mentioned there, I suspect it pretty
much just gave up.

Also, any dump program you give a symlink to is more than likely only
going to back up the symlink itself, not what it points to.

>Andrew A. Raines

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

Reply via email to