G'day all

I'm running NBU 4.5 on a 4cpu V880 with Solaris 8 server and we have a
number of classes that have True Image Restore enabled. These classes
are usually for filesystems with millions of files. The nbu database is
kept on a R5 T3. .

The problem is that when nbu does its housekeeping, after a scheduled
set of backups runs, it processes the on-disk TIR image info. (We had
set the retention of the TIR info to one day). For a client with a
number of backup images, each with TIR info for millions of files, this
can take a long time - i.e. a couple of hours. When you have 3 clients
with the same profile, this means a _long_ time doing the processing
housekeeping, which means in effect that no backups can run during that
time. Unfortunately, nbu does each client serially, one at a time. (Only
one or two cpus are kept busy, and the nbu database is on a Raid5 T3 and
i/o performance doesn't seem to be the problem).

I tried setting the retention period to 0, hoping that the TIR info
would only be kept on tape and not the disk, but it looks like it's
still going to disk.

Does anyone else have this problem, and do they have a workaround? Is
there something I'm missing?

Does nbu 5 do this housekeeping a bit smarter, with some degree of
parallelism, rather than one client at a time?

Is that the last question?

Thanks.

Marijan