We're currently at an standstill on how to resolve our NDMP restore problems
with TSM and Mirapoint and I'm reaching out to this list to see if anyone else
has experienced this problem or any potential solutions:

We have a series of Mirapoint mail filers running MOS 3.8.5-GA.
We're trying to configure 3-way NDMP backups and restores via TSM (NDMP version
4) on our TSM server version 5.4.2 on AIX 5.3.

We've managed to get backups to work with both TOC enabled and disabled.  The
problem occurs on restore of NDMP data (both full image dumps and incremental
via TOC).  Within 5 minutes of initiating a restore "restore node mirapoint
/usr/store" the restore fails out without any real discriptive message:

06/17/08   17:11:08      ANR1065I Restore of NAS node MIRAPOINT, file system
                          /usr/store, started as process 50 by administrator
                          SFUENTES.  A full image for this file system will be
                          restored to destination /usr/store.  (SESSION: 40,
                          PROCESS: 50)
06/17/08   17:11:27      ANR1104E NAS Restore process 50 terminated - NDMP 
session
                          errors encountered. (SESSION: 40, PROCESS: 50)

06/17/08   17:11:27      ANR0985I Process 50 for RESTORE NAS (FULL) running in 
the
                          BACKGROUND completed with completion state FAILURE at
                          17:11:27. (SESSION: 40, PROCESS: 50)

In this case it was a matter of seconds, but the same message repeats for any
restore.  There are no successful mounts of any tape drives, the /usr/store
volume is knocked offline and MOS has to be rebooted.

The kicker is that back in April when we were ready to deploy for production,
somehow all the planets aligned correctly and the restores actually worked for
2-3 weeks.  Once we moved the TSM instance to a production server however, the
restores failed once again and they've failed ever since.  Even moving back to
the dev server, restores still fail.

We've done TSM server traces, low-level network traces and we're about ready to
get Mirapoint support on the line.  An IBM call has been opened but has not
yielded much.  I've checked on relevant APARS and though some are similar to our
problem, the APARS have been fixed on 5.4.2 but the restore problem still 
occurs.

More info for those interested:

q datamover:
               Data Mover Name: MIRAPOINT
               Data Mover Type: NAS
                    IP Address: hidden
            TCP/IP Port Number: 10000
                     User Name: hidden
      Storage Pool Data Format: NDMP Dump
                       On-Line: Yes
Last Update by (administrator): hidden
         Last Update Date/Time: 06/12/08   12:53:15


on MOS:
ndmp get version
4
ndmp get dma
default
ndmp get preferredip
(same as IP Address above)

Is there something I'm overlooking? Any ideas?

Thanks!

Sergio
U. of Maryland