Hi John,

Thanks for your comments. I noticed the following runtar error in
sendsize.debug: "runtar: error [must be invoked by root]". The problem
is that for the rest of the machines I compiled amanda with user amanda,
while for the last SGI I used the precompiled tardist. The precompiled
version seems to be compiled with user root. The amanda server contacts
the clients as amanda, while the last SGI expects not amanda but root. I
wonder is there a way to change the user for amanda or I have to compile
amanda from source?

Thanks,
Ted

P.S. Here is sendsize.debug:

bash-2.04# more sendsize.debug 
sendsize: debug 1 pid 647466 ruid 1136 euid 1136 start time Tue Oct 24
13:11:42 
2000
/usr/freeware/libexec/sendsize: version 2.4.1p1
calculating for amname '/usr5/people', dirname '/usr5/people'
sendsize: getting size via gnutar for /usr5/people level 0
sendsize: running "/usr/freeware/libexec/runtar --create --directory
/usr5/peopl
e --listed-incremental
/var/amanda/gnutar-lists/craft14_usr5_people_0.new --spar
se --one-file-system --ignore-failed-read --totals --file /dev/null
--exclude-fr
om=/usr/local/etc/amanda/exclude ."
runtar: error [must be invoked by root]

.....
(no size line match in above gnutar output)
.....
sendsize: pid 647466 finish time Tue Oct 24 13:11:42 2000


"John R. Jackson" wrote:
> 
> >   craft14    /usr7/people lev 0 FAILED [disk /usr7/people offline on craft14
> >?]
> 
> You probably need to catch a sendbackup*debug file on that client that
> is for this file system.  There are three ways to do this:
> 
>   * Catch Amanda while it is trying to back up this file system on
>     this client and before sendbackup.debug gets overwritten with the
>     next backup (which is difficult).
> 
>   * Change your disklist to only contain this client/file system and
>     do a backup of just it so sendbackup.debug on the client only has
>     the output from attempting this one file system.  You might also
>     temporarily set "record no" and "tapedev no-such-device" so the
>     backup does not really do anything.
> 
>   * Rebuild Amanda with --with-pid-debug-files so each process on the
>     client has its own debug file, then you can find the one that matches
>     this failure.  If you do this, you'll have to remember to clean out
>     /tmp/amanda occasionally.
> 
> >Ted
> 
> John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]

Reply via email to