Basically, they claim it must be the hardware or windows drivers for the network card, because 'thousands of our customers have W2K and none of them have reported this problem'.
On the good side, the OTHER critical error was that we couldn't restore. anything. oops.
Turned out that if it was never merged into the database, BrightStor won't restore it. The merge was taking over 24 hrs and failing because the next backup began before it finished. It would usually catch up over the weekend, but some sessions were hosed, and thus no restores.
So we extended the database by 2 more chunks. Ingres, by the way, has a 2gb database limit. Thus we now have effectively a 6gb database. Did I mention that it took only 3 days to put that first database to 80% of capacity? Thus this 6gb database will last another 4-5 days before it too begins warning me that I am out of room. Oh yes, if it runs out of room, it won't merge the backup sessions.
Now, extending the database made the merge take a little over 10 hours instead of 24, but only because there was so much room to spare. It will quickly slow down. The solution is to either buy a 120gb drive and extend it WAYYYY out there, (which is impossible because it is installed on a sparc20, and it requires scsi drives, and can only hold two) OR stop detail logging (which means you have to merge the tape inventory into the database before you can restore anything) OR stop doing full backups of all 9 servers nightly, and move to a differential scheme. This latter is what we will do, most likely.
