Failed SLP and associated tape

[Amaan]

Master server NBU ver: 7.0.1

os: Win 2008 R2

Media server: 7.0.1

OS: win 2003

Storage: Quantum dxi4520 and tape library.

Backup jobs are running to Quantum DXi4520 (OST) and will be duplicated to tape. We are using SLP for this. Full image size is approximatly ~6.5 TB. Tapes are LTO2. Check point is enabled in the policy.

Question: When duplciation job was running only one tape available and it wrote 1.7 TB and failed with 96, as we had no tape. When this job retries does it begin from the point which was failed. If no what will be with the data (1.7TB)  which was written previously and couldnt finish? Is it possible to erase it?

[Mark_Solutions · Accepted solution]

It will start again from scratch All tapes previously used for this duplication will get the image fragments expired during the next image cleanup Duplication has no concept of checkpoints I am afraid If you run bpexpdate -deassign_empty you will see the tapes clear down (please check my syntax on that command as I am out of the office) Hope this helps

[Amaan]

Thanks for answer.

What is the difference between this command and between bpexpdate -m id -d 0. as i understand both of them will clear the tape assign tab.

[Mark_Solutions]

The command I gave just deassigns expired tapes immediately rather than waiting for NetBackup to do its regular cleanup A bpexpdate with a date of 0 expires images that are valid - the half written ones to tape are not, they are already expired - the cleanup command is totally safe as it just cleans up tapes that have already expired NetBackup runs this twice daily but can also run the image cleanup job at the end of a failed job to Rudy things up Hope this helps - your choice and quite often NetBackup has dealt with this before we even notice and the rerun of the SLP will usually start.overwriting those tapes anyway

[mph999]

bpexpdate -deassign empty will deassign and media that contains no images, but where the media still has as assign time.

The bpexpdate -d 0 -m <media id> will deassign the media,if it contains images or not.

Care must be taken with this command.

Martin

[Mark_Solutions]

Just wanted to point out that the -d 0 option is for all images on the tape, so if any succesasful duplications were also done to one of the tapes they would also be expired

[Amaan]

Hi Mark,

does it mean that if the tape has two images. one is successful and one is failed but written. and if i run deaasignempty command it will just cleans the part of the media which has a unfinished image. Do i understand it correct?

[Mark_Solutions]

Not quite If it wrote the good duplication up to block 20 on the tape and then started the failed one, when the cleanup runs it knows that it can start writing from block 20 again and overwrite the failed one only The is a second cleanup process command that also deals with these small chunks of failed images - that is the command that runs during the image cleanup jobs you see in activity monitor Can't remember exactly but it is something like nbdelete -cleanup -all_clients (just off the top of my head) The reason I said try the deassign empty command is so that you would see the tapes clear down but the nbdelete, or a combination of both will tidy it all up for you As I said, in general NetBackup does it all for you I haven't covered the first image on a tape failing before the second image running as it is unlikely to happen as NetBackup would have tidied up before the next write to tape

[Amaan]

Thanks so much. My question has been answered.

[Mark_Solutions]

Great - the thing with tape is that you cannot reuse blocks early in a tape - it doesn't usually apply to duplications but if a multiplexed backup failed the early blocks could not be reused and would be wasted until the entire set of images on that tape expired