Hi All

I have a customer that thinks he wants a monthly backup kept forever: cites
legislative requirements and will not be disuaded.  Current implementation
is to use a series of backupsets.  

The issue with this is that the backupset generation can only be started
when a drive is available - it will not wait for a drive.  Further, if it
gets bigger than one tape, and the server is busy then tape 1 is
dismounted, another waiting process grabs the drive and tape 2 cannot be
mounted, causing the backupset generation process to fail.  The solution so
far has been to break up the backupset generation of multiple nodes into
tape sized bites, but as data volumes get larger one of these jobs seems to
fail every month.  Also I now need to run an export of a large database
(2TB) and export seems to suffer from the same failing.  Is this new
behavior with 5.5?  I don't remember coming across it before.

I have tried to address this by setting up two device classes, exactly
similar except that one has mountlimit of (number of drives) - 1 and the
other has mountlimit of 1.  I hoped that directing the export/backupset to
that device class would reserve one drive for it.  Testing this afternoon
has shown that this is ineffective though I am at a loss to explain why.

Do others have this problem?  What work-arounds are there?  Long term I
hope to make the backupsets run server-to-server which will improve the
reliability and tape utilization, but for the present I am stuck.


Thanks

Steve.

Steven Harris
TSM Admin
Paraparaumu, New Zealand.