HELP on missing savesets and client indexes

Question

Hello to all,
Can somebody please tell me what might have happened or why might have caused this to happen?
I am trying to restore files from our month end tapes (5 months ago), according to the tape log we have these set of tapes, so I ordered them from offsite, loaded them in the jukebox and when I check what's inside the tapes - they are all blank/empty. I used 'mminfo -avot -c ' and that too didn't showed up any savesets for the month I'm trying to restore from. Manually loaded 1 tape from the month end set of tapes and started scanning using 'scanner -iv' and when done, I checked the tape and there savesets in there now and also mminfo is also showing the indexes.
I have this issue on 4 month end savesets or sets of tapes and that involves about 50+ tapes, so as much as possible I don't to use scanner on each of them, but is there another option to rebuild the media and client index from these tapes?

thanks in advance

ble1 · Answer

Without full scale investigation one can only guess what happened - and it may be virtually anything. Since mminfo didn't show anything it means those savesets were no longer recorded in media database. My question is, what exactly did you do when you said 'oaded them in the jukebox and when I check what's inside the tapes'. I really don't understand that part. Logical question is, is that volume part of media database as well? Well, now it is, but what about the others? If they are reported by mminfo as not there it is possible that whole volumes were removed from mdb. mminfo -avot -r volume | sort -u will give you the list of volumes inside mdb. Knowing if volumes are missing or volumes are there but save sets are missing is a starting point.

lpamintuan · Answer

the volumes are still on the list or in the mdb, but as I said when I use mminfo -avot N10403, there are no savesets showing up or missing, tape is showing blank but full.

lpamintuan · Answer

root@backupserv >#mminfo -avot -q 'volumename=N10403'6095:mminfo: no matches found for the quey

mgennacc · Answer

Hello,

I would also try running mminfo with this syntax, I have seen occassions where using the volume name with qualifiying it (-q volume=xx) did not produce output, but using the -q volume=xxx does.

what does this produce?

mminfo -avot -q "volume=N10403"

ble1 · Answer

OK, next step then. grep all your logs for N10403 to see if after last backup was written anything happened there.

If you have rap logging enabled also check it out.

If you have user audit in NMC enabled, check it out too. Save sets do not disappear like that so there must be some sort of trace somewhere.

lpamintuan · Answer

I've done already doen that and also used this command if ever by some chance somebody relabeled the tapes:

mminfo -avot -q Volume=VolumeLabel -r olabel,labeled

and they were not (last labeled date was before the tapes were used). I'm also using disk backup then month end savesets cloned to tapes. Tapes are immediately sent to offsite storage for 5 years. They are only recalled whenever we have file restore requests and the tape is also set to READ only (using the safety tab on the cartridge) to prevent accidental writing or relabeling. The only line I can find from the log file is showing
'completion notice writing to volume N10403'

thanks

ble1 · Answer

The line, where do you see it? Is it after tape was filled (marked full)? Obviously you would also need to see nsrd message which saveset is writing and when is done, eg:
11/06/09 10:47:20 nsrd hcrvelin.local:/stand saving to pool 'fsR' (TC200001)
11/06/09 10:47:22 nsrd hcrvelin.local:/stand done saving to pool 'fsR' (TC200001) 57 KB

If for some reason you see completion notice only and other barcode was used to save then you must recheck barcode vs volume name listing:
mminfo -avot -r barcode,volume | awk '{if ("" $1 != "" $2) print $1}' (or something like that)

lpamintuan · Answer

it is not showing any saveset, just the amount of data written to tape:used 250 GB of 200 GB.My suspect is that my media database got corrupted.

ble1 · Answer

Now that's strange. Well, I do not think that is mdb yet as when saving data you should see volume name too in log and that has nothing to do with mdb. Unless your barcode !=volume name. In that case, unless you are careful, you may get into position where you can create duplicate entry and at mdb level make full tape empty.

lpamintuan · Answer

Thanks a lot for help/advise Hrvoje Crvelin. I now have the answer on my issue/problem. The problem was caused by the user or myself (not intentionally though). I was about to do the same mistake again and that's when I realized why my savesets/indexes went missing.
We have a 9 TB disk backup device, but because of the amount of data and retention period (12 weeks), I am always low on disk space after 12 weeks. So what I was doing it to manually clone (not stage) the month end savesets to tapes, then delete the entire week off of the disks using the SSID. This is the issue with using disk backup, even if Networker marked the savesets as 'expired or recyclable', it will not remove it from the disks, unlike tape, you can re-label it, then get a full tape available for backups again.
The 9TB disk backup device with dedup is only partitioned with 2 NAS shares and can't re-label because I still have valid savesets in there.

ble1 · Answer

You should use ssid/cloneid when deleting savesets. As for disk, NW should remove saveset from disk once expired. If not, then you have another problem.

NetWorker

HELP on missing savesets and client indexes

Was this post helpful?