Tapes used on a failed job

[Leoooooo]

Good morning,

This weekend I had something new for me and I'd like to ask for some assistance of who has more experience about this. (NBU7.6)

Friday I left my library with 3 scratch tapes to run 3 jobs.

Today (monday) I could see that all jobs have failed. One of the jobs utilized 2 tapes and requested a 3rd one which there wasn't and another job failed asking for a second tape that there wasn't and a 3rd job that hasn't started because that was not scratch.

Ok, so far so good. The point is, when I check the tapes inside the library they all show as SCRATCH.... :|

What is so weird, because I can see they have been used. On reports these tapes doesn't exist (as they are scratch) the only way I could make nbu tell me that there were something on these tapes was when I tried to expired the images, then nbu tell me that there are images on the tapes and the correct expire date.

Again, so far so good. I found a doc explaining how nbu manages the tapes, but not quite how it works for FAILED jobs especific.

So my request would be: does someone has a document to understand how nbu works with FAILED job images on tapes? And I want to keep track of these tapes used this weekend on the correct volume pool and which images I have on them, how do I do it? If I need to use these tapes for restore, how to procceed as they show as scratch?

Thank you in advance.

[Leoooooo]

Trying to post

[mph999 · Accepted solution]

Yep, that's correct, it works like this ... If all backups jobs fail then the tapes are unassigned from the media DB and go back t scratch, why, because there is nothing on them of any use. If however, the tape had one good backup, and the 2nd failed, then the tape would rewind to the beginning of the 2nd backup so the space could be reused (that contaned the failed backup). If a job ran and was successful on one empty tape, then another job ran the spanned to a second empty tap, and then failed, the 2nd tape would be unassigned (again nothing good on it) but the first part of that bad backup on the first tape would be left (so we don't reload a tape to rewrite the logical-end-of-data mark in this secnario). Therefore the firrst tape would have some wasted space until it expires.

[Leoooooo]

Thank you Martin.

So for a job that took 2 tapes and failed requiring more tapes, these two tapes that are full of data, in fact, has no usable data in case of a restore? If these tapes has been reassigned to scratch, why aren't they scratch indeed as they have data on then? These tapes are scratch or not?

Thanks

[Marianne]

With failed jobs, NBU will discard all image info for those jobs. 

If there are no other good, successful backups on the tape, NBU will deassign the tape again after the failure. This will make tape available to be overwritten again.
Same as when all images on a tape have expired.

Try to break up backups in smaller jobs, e.g. 'Allow multiple data streams' will break up items in Backup Selection into individual streams, giving you more chance of having restorable data.

Have a look at this excellent blog: 
Understanding how NetBackup writes to a tape. 

[mph999]

So for a job that took 2 tapes and failed requiring more tapes, these two tapes that are full of data, in fact, has no usable data in case of a restore ?

Yes, they have no useable data, so NBU will return them to the scratch pool, providing they started in the scratch pool in the first place.

(If you manually move an empty tape from scratch to say, NetBackup pool, then run a job, when that tape expires it will stay in the NetBackup pool, only tapes taken from scratch by NetBackup itself, will return to scratch).

You stated:

If these tapes has been reassigned to scratch, why aren't they scratch indeed as they have data on then? These tapes are scratch or not?

I don't understand this bit " why aren't they scratch indeed as they have data on then"

If they have been assigned to scrach, then ther are scratch - simple as that ...