Greetings, all.
I'm working on a plan to accomplish the charge to back up a database such that
"No committed transaction is lost", in the case of a disaster. Beginning to
noodle around the corner cases of the process, it occurred to me that a SAN
environment could make for several interesting possibilities. For example,
how many of you:
- Put the various mirrors of your database volumes in different buildings?
- Mirror (at the OS level) storage pool volumes?
(of course, I'm talking here about different buildings too)
Would those of you who have sharks choose shark-y replication before you'd do
these other things?
The trick is that, as soon as TSM feels that it's finished storing a file, I
need to be confident that a single-site disaster will _not_ lose me the file
(for this particular class of file: DB/2 Transaction logs...) That means the
storage pool, the DB, the DB logs, everything needs to be at least a little
geographically diverse -right- -then-.
- Allen S. Rout
|