Veritas-bu

Re: [Veritas-bu] massive catalog growth

2011-11-14 21:56:44
Subject: Re: [Veritas-bu] massive catalog growth
From: "bob944" <bob944 AT attglobal DOT net>
To: <veritas-bu AT mailman.eng.auburn DOT edu>
Date: Mon, 14 Nov 2011 21:56:30 -0500
> Great, so it appears NetBackup is actually not
> respecting my exlcude_list. I just ran a test job
> and realize it is backing up DSSU's, the catalog
> and a whole bunch of other things that should be
> and are excluded.

And we all assume it's not _really_ named "exlcude_list", right? :-)

Not that I've seen every NetBackup problem by any means, but since
NetBackup 3.1, I have never seen {ex|in}clude_list processing not work
correctly.

Maybe this is that first time, maybe not.  Suggest you look at your
bpbkar log; it (at verbosity 1, IIRC) will show all of the *clude_list
processing, file by file.  When I don't understand why something is
included or excluded, that has always shown me my error.

Also, related to your catalog growth.  There's a technote or two
explaining the per-file catalog entry.  Often, each file is
generalized to take up 100-130 bytes (less with binary catalogs).
That number is a good working average, but can be way off if the
pathnames--almost always the longest part of each entry--are long.
Backups of Windows clients with a lot of filer shares mounted often
have paths over 400 bytes long, IME.  The size of the files file of a
backup of that sort of path could be five-or-more times larger than
the same number of files with normal-length paths.  FWIW, the one
Isilon box I ever dealt with had very long paths if backed up from the
root.

bpflist is a cantankerous command, but its 17 fields will clearly show
how much of the differences in your older backups and newer ones are
due to path lengths.  Even simpler, divide the size of the .f file by
the number of files backed up by that stream to compute your own
metrics on bytes-per-file-entry.

If you're still having difficulty, it could be helpful to post
platform and NetBackup versions of the master and clients involved,
our exclude list and bpbkar log entries that illustrate the problem.


_______________________________________________
Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu

<Prev in Thread] Current Thread [Next in Thread>