I found similar notes on the veritas support forums. Some documents say
that the actual file scan is performed AFTER bpstart_notify is run, so
some things can be done to exclude files or directories just prior to
the backup. Since bpbkar kicks off the bpstart_notify, and presumably
the last bpbkar must also run the bpend_notify, it seems I cannot make
NB dynamically create a file list at this time. I think I observed the
behavior you described. I launched the backup, and eight bpbkar jobs
appeared in the task list. The file scans began, and the system began
to stream data to the master server.
Thanks for your help.
>>> I'm working to maximize throughput from a Win2K system, which backs
>>> up our two NetApp systems through mapped drives. When I do this
>>> using one data stream per filer, it takes 66 hours, and the tape
>>> shoe-shines quite a bit. I have built a file list that breaks some
>>> top level directories into mostly equal "NEW_STREAM" data streams,
>>> so they all seem to finish within an hour of each other. This keeps
>>> the tape drives moving data nicely. I have written a Perl for Win32
>>> script that scans the top level of each of my NetApp servers
>>> (through the mapped drive), and outputs a file format that matches
>>> the format used in the configuration on the NB master server. This
>>> would work well in a static environment, but one of the NetApps
>>> contains user home directories. This changes sometimes without my
>>> knowing it.
>>> I don't see anything on the Veritas support page regarding the
>>> creation of a dynamic file include list that can contain the
>>> declaration of "NEW_STREAM". I would like the ability to launch the
>>> backup, and have bpstart_notify dynamically create an agressive file
>>> list, and begin the backup.
>bpstart_notify is launched by bpbkar which runs for each stream.
>Hence, the streams have already been created by the time bpstart
>>> Does anyone do this already?
>>> NB 4.5, Solaris 9. All media servers are sun. LTO-1 drives.
>>> I immagine it's possible to push the dynamic file list from the
>>> client to the master server, but that would expose the master
>>> server, and some very sensitive config files.
> -- Larry Kingery If at first you DO succeed, try not to look astonished!