Michael-
I would actually disagree with Veritas support on this one, I think
they got it backwards. :-/
Duplication is a pretty good way to validate that you've got a good
backup tape, but a bpverify IS the more thorough choice. There are
some rare situations (specific hardware misconfiguration at the tape
driver level) where an image can be duplicated whereas the bpverify and
restore operations fail. I don't think I've ever seen or heard of a
situation where a bpverify worked for an image but the restore did not.
If you care about the details as to why, read on... (sorry, I'm
feeling wordy tonight :-)
Suppose you have a NetBackup tape with two non-multiplexed images
written on it.
As far as the true contents of this tape (as displayed by the "tcopy"
command for example), this example NetBackup tape actually has the
following pieces of data written on it, in this order:
(file 1) the NBU media header
(file 2) an NBU backup header for image #1
(file 3) the tar file that is the actual data of image#1
(file 4) an NBU backup header for image #2
(file 5) the tar file that is the actual data of image#2
EOT
If image#2 is being DUPLICATED, NetBackup does the following:
1-mounts the tape
2-checks the media header
3-positions to file 4
4-checks the backup header
5-positions to file 5
6-reads from file 5 until an End Of File character is reached,
blindly duplicating each block of data
7-cleans up
But if image#2 is being VERIFIED, the above steps only apply until you
reach step 6.
Instead of blindly reading blocks out of file 5, the bpverify process
is going a bit deeper. It does the equivalent of a "tar tf" operation
against the tar data that makes up file5 while comparing that
information with what is stored in the NBU catalog.
So bpverify is scanning the tar file's table of contents, instead of
just blindly reading and writing each block like duplication does.
HTH
rob
On Apr 6, 2005, at 2:27 PM, Michael L. Barrow wrote:
> I'm trying to find out some details on what bpverify is testing when it
> reads images on media. My intent is to make sure that the media hasn't
> suffered some type of corruption and the stuff on the tape matches
> what's
> in the catalog.
>
> Specifically, I need to understand exactly what a successful bpverify
> means and how bpverify compares to bpduplicate. In short, is
> bpduplicate a
> significantly better test of an image written to a piece of media, or
> is
> bpverify "good enough"?
>
> The confusion is the result of a comment by a Veritas NBU support
> engineer that indicated that bpverify doesn't do as rigorous a job as a
> bpduplicate.
>
> This verification step is intending to be a quick spot check in
> addition
> to more intensive restore testing.
>
> Any thoughts or pointers on this would be appreciated.
>
> --
> Michael Barrow
> michael AT mlbarrow DOT com
>
> _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
|