There have been several posts recently about various aspects of
duplication. Here
is a description of the way we do things here which I hope others will
find useful.
To start with, let's make some observations about the way NetBackup
handles
the duplication process:
(i) Duplication jobs seem to be "invisible" to other activities
such as
backup and restore. For example, dup jobs don't show up in the
xbpmon screen.
(ii) Because of this "invisibility", you can get into trouble if you
have dup jobs running when you want to do a backup as the backup
process will think there are storage units / tape drives
available
when in fact there are none (and so you'll get a media mount
timeout
causing the job to re-queue -- as long as the dup run finishes
during
the
backup window, the backup job will complete but you will, of
course,
get
lots of error messages / emails which is not very satisfactory.
(iii) If you had multiple (say 2) libraries, you could mark the S/U
representing
one of them as "On Demand Only". Then, assuming your class(es)
use
"Any
Storage Unit", they will skip this one and not run into the
problems
of (ii).
(iv) The trick here is to find a way to "logically" subdivide a
library
into
multiple storage units so that multiple dup runs can take place
at the
same time.
After lots of head-scratching and calls to the Veritas support desk, the
recipe we
use is (starting with a 4-drive ATL P3000 library configured as a single
S/U):
(a) Change the S/U S0 to have 2 drives instead of four
(b) Create two new S/U's, each with a single drive. Make
the
drive type be DLT2
(c) In device manager, stop the daemon then use Update Drive
to change the density of two of the drives to be DLT2
and
then restart the daemon
What you then have is:
S0 contains 2 dlt drives
S1 " 1 dlt2 drive
S2 " 1 dlt2 drive
Now S1 and S2 can both be specified as "destination storage
units" in
the
Images->Duplicate panel, i.e. both can be active at once. So,
our
procedure is to generate the list of backup id's to be
duplicated and
then select half (ish) to use S1 and the rest to use S2.
This, of course, will use up all 4 drives (since the 2 in S0 are
needed
to read the originals) ... for this reason, we're adding a
fifth drive
which can be used for backups / restores while duplications are
in
process.
Let me know if this isn't clear.
Cheers,
Richard.
|