I've been talking to some of our Oracle DBA's, and we have discovered
that we have a problem.
We've been using SQL*Backtrack with Oracle for about the last 9 months
during the commissioning of a complex system for UK Electricity
deregulation. What we now have is a primary storage pool which is about
100 3575 tapes (about 1.5TB) of largely useless test data, together with
two offsite copy pools of about the same size.
Unfortunately, what we have found is that SQL*Backtrack names each table
backup with a constructed name that includes the date that the backup
was taken. This means that each backed up entity has a unique name, and
therefore that normal versioning does not kick in. As we keep the last
version of a file forever, this means that nothing is ever deleted.
Due to a tremendous lack of foresight, we are using the same management
class as our normal dsmc incrementals for said machines. This means that
I cannot just turn on ageing of the data without affecting the other
backups that we have.
Again, with more thought, I would probably have deleted or renamed the
filespaces that were created before we went live with the systems.
I also have a problem in that query filespace on the filespaces that
SQL*Backup has created shows the filespace but shows the amount of data
backed up as 0.0K. Now this could either be because the data is too
large, or that the normal tools cannot handle SQL*Backup filespaces. I
think that the latter is the case, as I cannot see the contents of the
filespace using dsmc on the client, even if I use the correct tag. I
know that there is data, because a query content on tapes selected at
random shows entities (I shall refrain from calling them files) in the
SQL*Backtrack filespace.
My DBA colleagues do not appear to be able to see a way of deleting the
table backups in SQL*Backtrack (nor even referencing them if we have
lost the profiles!).
Now, the challenge is to delete some of the useless data (almost all
SQL*Backtrack backups until about a week ago) without affecting any of
the other backups. There does not appear to be a way in SQL*Backtrack,
and I have missed the window of opportunity to do it using filespaces.
Anybody any ideas? My server is on AIX 4.2.1, and is at 2.1.5.18. The
client code is on AIX and Digital Unix, and is 2.1.0.8 (I think on all
platforms), and we are using version 2 of SQL*Backtrack.
Any help gratefully received.
Peter Gathercole
Peter.Gathercole AT virgin DOT net
|
