Not that I have found...

     BUT !

     perform your exports one file system at a time, then if/when it fails
     you waste less time !

     On your import you may then import only the specific file system's
     data and use the "replacedef=yes" flag and not have any impact on the
     already imported file systems.

     Let me give you some added info while we are on the subject...

     to get the import to work you have to define a server,  on the
     destination server, of itself BUT use the nodename=theotherserver...
     then define the device class just as it is on the origional source
     server.

     if you hit a snag send an urgent mail message directly to me and I'll
     assist (try at least... sometimes what I spew out isn't much help :-()
     I'm on call so I'll be checking my mail over the weekend also...

     I've done TONS of this and yes, clients with over 50 GB's of data I
     usually end up doing them one file system at a time.

     another hint... when you defined your device class what did you give
     its extimated capacity ? ? ?   I  generally define mine at 1024G this
     way there are fewer virtual volumes for the adsm systems to track.
     Even with this, as soon as a task finishes and closes the virtual
     volume it marks the virtual volume as "FULL" so it doesn't try to add
     info to it with a later process.

     later,
           Dwight


______________________________ Reply Separator _________________________________
Subject: server-to-server export fails on tape write error
Author:  j-valdes (j-valdes AT UCHICAGO DOT EDU) at unix,mime
Date:    2/12/99 12:25 PM


All,

I'm trying to export a node from one ADSM server (AIX 4.2.1, ADSM
v3.1.2.1) to another (AIX 4.2.1, ADSM v3.2.1.0) using server-to-server
transfer.  I've configured both servers, and the source server is able
to export its data to the target server.  The archive pool on the
target server being used to store the source server's data uses 3590
tapes in a 3494 tape library.  I'm having a problem, however, whenever
the target server encounters a media error when writing to a 3590 tape
in the destination archive pool.  When this happens, the target server
marks the tape as read-only (which is fine) but terminates the export
process (which is not fine).  I then need to restart the entire export
from the beginning.  I need to export close to 1TB of data for this
node, so having to restart the whole export again is quite a waste of
time.  I'm able to transfer about 5MB/s between the servers (they're
connected via ATM, and the pools are setup so that the transfer goes
from tape on the source straight to tape on the target), but even at
that rate, it will take over 52 hours to complete the transfer as is.

Is there anyway to tell the target/source server to just move on to a
new tape whenever it encounters a write error and not terminate the
export process?  Alternately, is there a way to tell the source server
to restart the export process from where it left off, rather than
having to restart it from the beginning?

Any suggestions would be appreciated.


John

-------------------------------------------------------------------------
John Valdes                        Department of Astronomy & Astrophysics
John Valdes                        Department of Astronomy & Astrophysics
j-valdes AT uchicago DOT edu                               University of Chicago