Many thanks to all those who responded to my questions about backup. In summary it would seem that if I'm going to have 500GB filesystems then nothing hokey is going to cut it - I need something to either manage a locally attached bunch of tape drives and an autoloader.
The only viable alternative seems to be either dumps across rmt (or rsh) or NDMP to a host with a large bunch of tape drives in a larger library.
Our enterprise backup product doesn't support any of scenarios yet. I need something cheap and cheerful to manage this for me. The dumps to tape are only for a complete filesystem recovery, we would run snapmirror to keep the users going while rebuilding a broken box. File level stuff I'll handle through a CIFS mount on an NT box where our backup client will run.
This leads to 2 questions 1) Which cheap and cheerful product to take care of the dumps and which of the ways above? Remember the dumps are only for volume recovery.
2) Has anyone worked out what the right way to failback after cutting over to a snap mirror replica? I understand netapp are working to make it easier but right now it looks close to unworkable with large volumes of data. If there are clever answers here then would anyone take the risk of not having a volume level dump?
TIA, Al