hend_bs

updated

5 years ago

H

hend_bs

1 Rookie

•

34 Posts

0

2034

June 12th, 2021 09:00

issue scanning a tape volume

Hello,

 

I need to make an SSID browsable so to rebuilt the indexes I am using scanner -i command against the tape volume.

first problem I faced is LTO Ultrium version conflict (drives are LTO 7 and the tape is lTO 5) I fixed the issue by deleting the volume from mm DB  (I kept a copy of old mm)using nsrmm -d command against the volume.

next I am receiving this error

[serviceBasis@SN~]$ sudo scanner -s lnxvib120 -i -S 1240990337 rd=SN:/dev/tape/by-id/scsi-35901b0e2223390d3-nst
8909:scanner: using 'rd=SN:/dev/tape/by-id/scsi-35901b0e2223390d3-nst' as the device name
8936:scanner: scanning LTO Ultrium-7 tape 102368L5 on rd=SN:/dev/tape/by-id/scsi-35901b0e2223390d3-nst
32350:scanner: adding LTO Ultrium-7 tape 102368L5 to pool MS Clone1
8753:scanner: unexpected file number, wanted 66 got 2
8761:scanner: done with LTO Ultrium-7 tape 102368L5

(I tried with all available drives of the library) I ve been reading answers on similar issue but it was back in 2010, that are mentioning that it could be a reset of the drive and data is probably lost.. I am concerned.. 

  • bingo.1

    1 Rookie

    •

    2430 Posts

    1490

    1

    Posted June 13th, 2021 07:00

    First of all - you will only loose data if you (accidentally) relabel/recycle a volume. Hope you did not do that. To prevent this, you might also use the write-protect notch on the cartridge itself.

    I wonder if you can use "scanner -i -S ssid..." right now because by deleting the volume from the media db you deleted the reference of that ssid and NW cannot directly resolve the ssid to a physical location on tape (file or tape mark and record number). So load the cartridge into the drive (NW will not be able to mount it yet) and run "scanner -m ..."

    The block size should not matter - NW will adjust that by itself. The LTO drives will be able to read 2 versions back. Just make sure that the drivers will cover all versions.

     

     

  • bingo.1

    1 Rookie

    •

    2430 Posts

    1461

    0

    Posted June 16th, 2021 02:00

    Of course, a save set could spread over more than one tape, indicated by

      - h   for "header"

      - m  for "middle part" (maybe more than one)

      - t  for "tail"

    Unfortunately, if you delete the save set/volume info, you will loose this info as well.

     

  • hend_bs

    1 Rookie

    •

    34 Posts

    1485

    0

    Posted June 14th, 2021 07:00

    hello,

    yes I figured it out so I tried to run scanner but I run it without options and against the whole volume and the same error appeared and the command stopped.

    I will try to run it using -m switch and see

  • bingo.1

    1 Rookie

    •

    2430 Posts

    1478

    0

    Posted June 14th, 2021 08:00

    Add ad least 2 "-v" for added verbosity and pipe the output to a text file.

     

  • hend_bs

    1 Rookie

    •

    34 Posts

    1464

    0

    Posted June 16th, 2021 01:00

    Hello again,

    I have used " -im " switch, and the data became browsable.

    Also I want to mention that apparently scanning the tapes in order of the flags "hr, mr, tr" is important. is it true?

     

  • hend_bs

    1 Rookie

    •

    34 Posts

    1443

    0

    Posted June 19th, 2021 09:00

    hello again,

    In my case I marked needed tapes before deleting them from mm folder:

     ~]$ mminfo -avot -q "client=client" | grep "/vol001"  

    102348L5       LTO Ultrium-5 client 10/07/2016 10:16:01 PM 1700 GB 1240990337 mE full /srv/exports/vol001

    102365L5       LTO Ultrium-5 client 10/07/2016 10:16:01 PM 2918 GB 1240990337 tE full /srv/exports/vol001

    102368L5       LTO Ultrium-5 client 10/07/2016 10:16:01 PM 1033 GB 1240990337 hE full /srv/exports/vol001

    right away after making the save set browsable, I was able to browse with the command "recover" from CLI , (but at that moment I did not resume the recover job because I had to ask again which folder to restore exactly):

    but the good thing that I have copied the output before exiting:

    recover> cd vol001

    recover> changetime 10/07/16

    6497:recover: time changed to Fri 07 Oct 2016 11:59:59 PM CEST

     

    However next day I tried same command and the changetime returned:

    recover> pwd
    /srv/exports/vol001
    recover> changetime 10/07/2016
    6496:recover: cannot browse /srv/exports/vol001 as of `Fri 07 Oct 2016 11:59:59 PM CEST'

    I have made no changes and the save set is still browsable:

    [serviceBasis@lnxvib120 ~]$ mminfo -q ssid=1240990337 -r volume,cloneid,sumflags,ssretent
    volume clone id fl retent
    102348L5 1480640050 mb 09/18/2021
    102365L5 1480640050 tb 09/18/2021
    102368L5 1480640050 hb 09/18/2021

    Can any one explain this behavior and how to solve this?

     

  • bingo.1

    1 Rookie

    •

    2430 Posts

    1436

    0

    Posted June 19th, 2021 19:00

    Well - where shall I start? - It seems that you are caught by the different NW time formats...

     

    Usually, "savetime" or "savetime(#)" is the format which is understandable/readable for the user/admin. Unfortunately, it is not precise enough.

    "nsavetime" would be much better as it uses the UNIX/Epoch time, counting in seconds from 01/01/1970. This time stamp is precise but not 'understandable' for the user/admin.

    Just as an example, run a command like "mminfo -q= -r ssid, savetime, savetime(25), nsavetime, cloneid " and compare the output. First thing what you will notice is that nsavetime and cloneid will show the same numbers      Of course all parts of a save set will have the same time stamps.

     

    Now, if you use "changetime date" to define a certain savetime, this is unprecise because NW does not know which exact time you want. Therefore it 'auto-corrects' your input and includes ALL seconds for that day. This results in "date ... 11:59:59PM" to cover the whole day. This is exactly what you see. Of course you could also say "changetime= " or "changetime= " which will set the pointer to the exact second.

     

    btw - if you want to get rid of the (potentially huge) file index for a save set, you have 2 options:

      -  nsrmm -o recyclable -S ssid/cloneid                                    .... but

      -  nsrmm -o recoverable -S ssid/cloneid                                .... simply does not exist. For that purpose, you must 'purge' the save set with the command "nsrmm -dP -S ssid" and wait until the index will be cleaned by NW automatically or just run "nsrim" for all/that specific client.

    Clear as mud?

     

  • hend_bs

    1 Rookie

    •

    34 Posts

    1409

    0

    Posted July 1st, 2021 03:00

    Hello,

     

    Thanks. Crystal clear !

    However in my case nsavetime and cloneid have different values.

    I managed to find the tape that contains the index and nsrck -L7 worked and made data browsable.