Veritas-bu

Re: [Veritas-bu] "Successful" NDMP Backups Missing Data

2008-03-15 04:04:26
Subject: Re: [Veritas-bu] "Successful" NDMP Backups Missing Data
From: "bob944" <bob944 AT attglobal DOT net>
To: <veritas-bu AT mailman.eng.auburn DOT edu>
Date: Sat, 15 Mar 2008 03:47:07 -0400
> Client: Sun StorageTek 5370 (nbnfs) [...] 
> If you notice, all of the red entries are less than they 

Just FYI, there is no "red" in an ASCII mailing list.  But you included
enough context to see what it was that concerned you.

> (this one and another) are showing this same issue.  One
> intresting point about this particular host is that the
> same directories are missing from all three "successful" [...]
> We've had issues in the past with super long directory
> paths on this NAS device, maybe this is simmilar.

>>From the NetBackup Administrator's Guide for NDMP, "NDMP operating notes
and restrictions"

  o  You cannot back up files where the path length is
     greater than 1024 characters

Perhaps this is applicable?

> We've got a ticket open with Sun, as Netbackup seems to
> think its a filer issue.  (We agree.) [...]
> Personally, I'm appauled that there seems to be no better 
> communication between the filer and Netbackup about the
> backup status and that NBU just assumes everything's 100%
> hunky dory.

This makes _no_ sense.  1) you think it's a filer issue.  2) the filer
is returning a success status.  3) NetBackup is telling you the filer
said it was successful.  4) You're appalled?  "there's no better
communication" and "NBU just assumes everything's 100% hunky dory."  

You're running an *NDMP* backup.  THAT HAPPENS ON THE FILER.  The filer
returns success.  Jeez, mon, what on earth do you _want_ NetBackup to
do?  :-)  www.ndmp.org has the NDMP specs, assuming your Sun NDMP
implementation is compliant (inserted here for others' benefit; I know
you know this).

FWIW, I've _always_ charged my Ops staff to do is use their brains for
this very thing.  As in "if the 10-hour, 500GB Oracle instance finishes
with a stat 0 in five minutes and backs up 32KB, it's much more likely
that some bonehead DBA screwed up the script or the database than it is
that it really is a 32KB database today.  Go find out."

> I think its unreasonable to assume the storage guys need to
> check their filer logs every week to make sure their backups
> are complete.  

Curiosity:  What _does_ the filer show in this case?

> This may be an excuse for me to write the script I've been
> planning that measures job success on how much data was
> backed up (compared to the trend.)

Never really consdered this for a constantly growing environment, but
it's not a bad idea when the percentage of false positives is low.
Might be something that NOM/VBR/Aptare/* could implement, too.

> PS: Should have bought a NetApp =P

:-)


_______________________________________________
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>