Our backups ran fine this morning, so it would appear that it was the
recreation of the devices in NetBackup (following a boot -r last week)
that fixed the problem. I'll mail the list again if I ever find out
exactly why that particular fix was necessary.
Thanks to everyone who responded!
Justin
Justin C. Lloyd wrote:
> Ok, a backup just ran fine and ejected the tape, exactly as it should have.
>
> On Friday, since I had replaced the hardware a few times and done a boot
> -r during the past 2 weeks, I decided that it might be worth it to try
> deleting and recreating the drives and storage unit within NBU.
> Afterwards, I started getting status 96 errors. It turned out that when
> I recreated the stu, I created it with media type 8mm2 while the media
> was type 8mm. Once I changed the stu and the drives to 8mm, the test
> backup ran fine. So I'm going to let backups run tonight and see what
> happens.
>
> According to the Veritas support person I was working with this morning
> (the person from last week is out until Wed), the storage unit should
> not have to be modified or recreated when the hardware changes, at
> least, when the hw specs are the same as before. So neither of us are
> sure why it seems that recreating the stu fixed the problem.
>
> I'll send another update to the list tomorrow or wednesday once I'm sure
> the issue has been resolved. In the meantime, if anyone has any insight
> as to why that change may have been the fix, please let me (and the
> list) know. Thanks!
>
> Justin
>
>
> Justin C. Lloyd wrote:
>
>> I have a case open with Veritas (and Spectra Logic) on this, but so
>> far we haven't come up with anything and I'm getting desperate.
>>
>> I have a Spectra Logic 2K with 15 slots and 2 AIT-3 drives, and I'm
>> using NetBackup Datacenter 4.5 on Solaris 9 (64-bit). Things have
>> been running relatively fine, but about a week ago I would come in to
>> find that the catalog backups had not started because the two drives
>> were still occupied with tapes from backups. The backups had
>> completed but the tapes fail to eject. It looks like they're trying
>> to, because they start moving towards the drive doors, but then settle
>> back in. This repeats for about 30 seconds and then stops. The ONLY
>> way to get the tapes out is to physically remove the drives from the
>> library and manually extract the tapes. This happens EVERY time NBU
>> tries a backup.
>>
>> Now, if I use robtest, I can move tapes in and out of drives with no
>> problem. I cannot, however, use robtest to remove tapes inserted by
>> NBU. According to Veritas, NBU removes the tapes from the drives
>> using tpunmount, which uses /usr/bin/mt, which uses the OS SCSI tape
>> driver, /kernel/drv/st. On the other hand, robtest uses Veritas'
>> generic SCSI driver, /kernel/drv/sg.
>>
>> Now the *really* strange thing is that I can insert a tape with
>> robtest and remove it with mt.
>>
>> Note that the tapes are not *physically* stuck/jammed in the drives.
>> I can manually unload them without difficulty. So it would seem that
>> something about how NBU is mounting the tapes is causing this.
>> Everything used to work fine, this just started happening one day. I
>> have replaced hardware a couple of times before this started, but that
>> was generally due to actually hardware failures. I just installed a
>> new library last Monday, but that had no effect on the problem.
>>
>> I've set VERBOSE = 5 in bp.conf and added some touch files under
>> /usr/openv/volmgr for more logging output, but the Veritas engineer
>> has yet to find something.
>>
>> Anyone have any thoughts on this?
>>
>> Justin
>>
>
--
Justin C. Lloyd
Unix System Administrator
MCI System Technology Solutions
Office 703.886.3219 Vnet 806.3219
|