Veritas-bu

Re: [Veritas-bu] Dealing with frozen media in Netbackup

2010-06-16 18:52:01
Subject: Re: [Veritas-bu] Dealing with frozen media in Netbackup
From: Wayne T Smith <WTSmith AT maine DOT edu>
To: VERITAS-BU AT mailman.eng.auburn DOT edu
Date: Wed, 16 Jun 2010 18:51:55 -0400
I used to do more, but recently (on my v5.1 NetBackup setup) I've
  • look at /usr/openv/netbackup/db/media/errors once or twice a week, generally Monday and Friday when I tried to review that each client was having successful backups.
  • From the errors file, I'd look to see if there were "too many" errors or if there was a pattern of errors such as on one tape drive.  Clean that tape drive (cleaning may not be an appropriate action for your drive)
  • For each tape that had an error, freeze it:
    • /usr/openv/netbackup/bin/admincmd/bpmedia -freeze -m <tapeid> -h <serverid>
    • Note that if no data was written, the tape is free and can't be frozen. For this case, change the tape to a different pool, my Baudelaire pool (for tapes that have had a hard time).
  • Run the "Media Summary" job in verbose mode.  For each tape listed as frozen and expired:
    • expire it with "sudo bpexpdate -d 0 -force -m <tapeid>
    • change it to my Baudelaire pool
  • Now and again, whenever I was short of tapes or had the time or inclination, and a free tape drive, do a quick erase on the tapes in the Baudelaire pool and return the successful erased-tapes to Scratch.
  • Now and again, run the errors file through a stats program to find the number of errors for each tape.  Discard or return to vendor tapes with "too many" errors.  Alternatively, I'd sometimes suspect a tape of being "bad" and would run "cat /usr/openv/netbackup/db/media/errors|grep <tapeid>" to see the count, with the advantage of seeing the dates of each error.
I've never seen a tape automatically go out of freeze state, but I've frozen many more tapes than NetBackup does with its "x errors in y hours" rule.  Also, most of my "real" errors on my LTO-I tapes were write errors; almost all read errors could be accounted for by the tape recently having a write error.

Nicely, I won't be doing any of this soon, as I'm well into switching to a 6.5 system writing to a Quantum 6550 which has a remote 6550 partner.  I'm a big fan of writing to disk and using OST and a Storage Lifecycle Policy instead of tapes and "NetBackup Vault" to make copies of tapes.  Very slick.

Cheers, Wayne
_______________________________________________
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>