Forum Discussion

jim_dalton's avatar
jim_dalton
Level 6
11 years ago

Sanity check: never requested a clean?

NB7601, Solaris master/media svr plus windows 2008r2 media server. The latter has 2 drives for doing ndmp backups and two for non-ndmp and is the one I have a question about. The drives are hp lto4 fibre, in a HP msl robot.

In over three years there has not been at any time a tapealert to clean a drive. I'm struggling to believe this can be real. The drives are heavily used overnight and at weekends yet not a single clean. Tapealert is enabled, there is a cleaning tape (but of course its never been requested), and I've run the robot with auto-clean both off and on ( more out of curiosity : I believe it should be off ). My other media server gets through cleaning tapes regularly though there are 10 drives in my other robot and they are IBM LTO4: the loads are similar but the 10drive robot is a couple of years older.

There are tapealert msgs in the bptm log: process_tapealert / tapealert_and_release /  0x0 0x0, which suggests its all working fine.

Your thoughts and experiences appreciated. It would be nice to be able to trigger a tapealert just to check it does do as its supposed.

TIA,Jim

Backing Up
Backup and Recovery
Configuring
Managing Backup Devices
NetBackup
Solaris
  • Nicolai's avatar
    Nicolai
    11 years ago

    According to this link below you may have a optimal installation where the tape drives run at designed speed, thus no need for cleaning. 

    LTO drives are designed to remove normal levels of contamination automatically. Small internal brushes sweep the debris away before it can build up and become a problem. So many LTO drives never require supplementary cleaning with cleaning cartridges.

    http://docs.oracle.com/cd/E35103_07/en/E24606/html/get-started.htm

  • mph999's avatar
    mph999
    Level 6
    11 years ago
    Opps, just read your post again ... If NBU is cleaning, yes library should be off. If you do swap them round so library cleans, what happens ... 3 years is a long time, though I would have expected to see the drives cleaned well within that time. That said, if they haven't been cleaned and needed to be, I would also have expected them to quite simply stop working.
  • mph999's avatar
    mph999
    Level 6
    11 years ago
    I'll look to see if there is anyway to see what the tape drive is doing, scsi_command might help, it can dump the log from the drive itself - not sure how understandable it will be though. The only way you will get to the bottom of this, is if you log a call with the hardware vendor. Cleaning is nothing to do with NBU - and by that I mean NBU only cleans the drives when told to. I've never heard of a bug causing this, and given you receive other tape alerts I think that rules that out. Either way, you would need to start the t/shooting at the tape drive side, beacuse 'that's where the story starts', so to speak.
  • Nicolai's avatar
    Nicolai
    Moderator
    11 years ago

    According to this link below you may have a optimal installation where the tape drives run at designed speed, thus no need for cleaning. 

    LTO drives are designed to remove normal levels of contamination automatically. Small internal brushes sweep the debris away before it can build up and become a problem. So many LTO drives never require supplementary cleaning with cleaning cartridges.

    http://docs.oracle.com/cd/E35103_07/en/E24606/html/get-started.htm

  • mph999's avatar
    mph999
    Level 6
    11 years ago

    Just had a look through the code.

    It seems, though as you will image it is quite complex, that there is only one funcion that is used to read the tapealerts from the drive.  So if you're receiving 'some' alerts, it's working.

    In nutshell, the drive 'stores' the tape alerts, and NBU reads them, then clears them, on drive unload.

    So, if we can read one, we can read all ...

    scsi_command -d <drive path> -log_dump 

    .. will show the tapealert flags

    Eg

    ***TapeAlert Page (0x2e) -- ONLY flags that are SET

     

    (I don't have any set though)

    Guess (and this is a guess...) you would have to trigger the flags (if they're going to set) by using the drive outside NBU, else on unload they are going to be cleared leaving nothing for scsi_command -log_dump to report.

    So tar / cpio  / dd time ...

    ... though, given, as mentioned the same function reads the tape alerts each time, I think this is barking up the wrong tree.

    Vendor time me thinks, it's going to be quicker, even if you have to spell out tapealerts are outside NBU ...

     

  • mph999's avatar
    mph999
    Level 6
    11 years ago

    ... and having just seen the thread by the fabulous Nicolai, seems you might not have an issue ...  I didn't know the self cleaning was that good ...

  • Yasuhisa_Ishika's avatar
    Yasuhisa_Ishika
    Level 6
    11 years ago

    If you still have a doubt, check logs on MSL or try diagnostic tools(if provided by HP) on each tape drives.