If you've ever used VxFS's filesystem snapshot capabilities, it's very
much like that.  You define a slice (with no filesystem needed) as your
"cache" device.  This is where all of the "changes" to the filesystem get
stored while the snapshot is active.  This would be /dev/rdsk/c0t0d1s0,
in your case.  It shouldn't matter if /dev/rdsk/c0t0d2s2 is mounted or
not, or if it is ufs or vxfs.  If it's not mounted, however, then you're
sure that your cache device won't fill up.

One thing I would look for is why you are getting the "open" failure of
the device.  If you run 'bperror -problems -client <client>' you should
be able to see which process responded with the error 12.  Then, make
sure /usr/openv/netbackup/logs/<procname> is created on the client...and
put VERBOSE = 9 in your bp.conf file on the client.  This should give
you enough logging to tell you why the file couldn't be opened.

I've seen errors in the past where the "cache" device you specify
fills up with changes, and it disables all backups using that device.
But those are normally reported as error 13's.  It looks like you aren't
even getting that far.

One other thing (which I wouldn't think would matter) is to make sure
your cache definition uses all caps: CACHE=/dev/rdsk/c0t0d1s0.  I don't
know if it's case sensitive or not, but it's an idea.

Todd

On Sun, Mar 12, 2000 at 02:57:25PM -0800, Monte Ambrose wrote:
> All,
> 
> I have a couple flash backup installs to do this week and have only done
> this
> (not quite succesfully) in a test environment.  I would feel a bit more
> comfortable
> if I knew what I did wrong.  Let me know if you have any ideas.
> 
> Environment: Sun Ultra 5 with 2.6.  NBU 3.2 and Flashbackup installed.
> Backing
>              up to an ATL robot with dlt5000 drives.  Regular file backup
> works.
> 
> Setup:       Class points to the same storage unit and volume pool as
> regular
>              backup class that works. 
>              class type is set to flasbackup
>              schedule is full,daily,12:00 on Sunday for 2 hour window
>              files
>                   cache=/dev/rdsk/c0t0d1s0
>                   /dev/rdsk/c0t0d2s2
>              client is the local server client with a cache file system
> defined
>              on /dev/rdsk/c0t0d1s0 and a filesystem to backup defined on  
>                 /dev/rdsk/c0t0d2s2
>              start a manual backup from jnbSA GUI.
> 
> Error:       backup is active and tape is mounted in drive.  I get status
> code 12
>              "An open of a file failed".  Doesn't seem like a valid code
> since I
>              am reading from a raw device.
> 
> I have tried this with UFS and VXFS file systems.  I have tried with the
> cache 
> file system mounted and unmounted.  The flashbackup manual indicates that it
> should
> not be mounted. Should the cache be a disk or slice with no file system
> defined?
> Any suggestions would be appreciated.
> 
> 
> Thanks,
> 
> Monte Ambrose
> Senior Consultant
> 
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu