HP DLT/80 tar/dd backup I/O Errors

Justin Bennett Justin.Bennett at dynabrade.com
Thu Sep 26 09:32:18 EDT 2002


Darin Perusich said:
> these tapes work for about a week for so in the new drive then they
> start to fail. have you tried using these tapes in a different drive?
> that could be a double edged sword, if the drive has a problem tapes
> used in that drive can pass those problems on to other drives.

They are brand new tapes, they won't work in the other dlt drives
according to the manual, and in practice.

>
> it could also be the scsi bus it self. is there anything else on the bus
>  that is throwing errors? i'm going to guess that you've checked the
> scsi  id's and nothing is on id #7.

Nope 0,1,2.

>
> i've never heard of having to demagnetize a tape before using it. it's
> possible that the dlt is writting something to the first block of the
> tape and when you dump to the tape that block is getting written over
> and causing the problem. just a thought.

I haven't heard either, but according to HP this drive does something to
the tape. I can't get it to be consistent, it happens at diffrent times,
sometimes it's 7 minutes in others 1 hour 30. It almost seems like it
can't keep up for some reason. It's on the same bus as the other drives so
their all running at the same speed scsi bus. Maybe the drive is faster
than the others? It's an NFS share but the network should keepup with the
SCSI tape drive. Can I get tar to dump better errors??


>
> since you have a number of dlt's have you ever looked into using amanda
> to dump you systems? you can string all your dlt's together and make a
> virtual changer to increase capacity.

I've used a number of things BRU for example, was nice for the tape
indexing stuff, but I hap problems with the database getting corrupted. I
guess I just don't trust these software, like to know that when I do a tar
it's gonna work, not relying on some database backen to catalog where my
stuff is. I've heard good things about Amanda just haven't played with it
yet.


This drive is just pissing me off, I should have just bought another one
of my surestores. But I was lurred in by the sleek 1U design and 4 bay
chassis.. :)

>
> Justin Bennett wrote:
>> Ok I just bought another HP DLT 40/80 GB Drive. This one I got a nice
>> 4 bay rack mount array for it and everything. I have two existing
>> 40/80 DLT drives (diffrent model) that work great. However I just
>> learned that any tape that was used in my old drives won't work in the
>> new drive, unless I demagnetize them (from the manual of the drive) It
>> was kicking my existing tapes out opon insertion. This is strange
>> because I didn't think there was any type of formating on tapes, but
>> there must be some kind of
>> initilization on the tape.
>>    Anyways thats not the issue, I bought a new set of tapes for this
>> drive
>> and everything was working fine for a few days.  Now I get Input
>> output errors when writting to the tape. I use tar and dd to set the
>> block size.
>>
>> tar -cf /dev/st0 proj --blocking-factor 128
>>
>> This is backing the proj directory. I get
>>
>> Thu Sep 26 00:00:10 EDT 2002
>>
>> Tape 0 Projects
>> tar: /dev/st0: Wrote only 0 of 65536 bytes
>> tar: Error is not recoverable: exiting now
>> Thu Sep 26 01:23:37 EDT 20021
>>
>> Sometimes it runs as few as 7 minutes in this case about an
>> hour and a half. When I come in The clean/media light is blinking, I
>> ran the cleaning cartridge several time, I still get the same thing
>> the next night.
>>
>> Any Ideas? Is an HP vs80 or something while the others are hp
>> surestores, the surestores work just fine. They are all on the same
>> SCSI bus, but the backups fire in sucsession, when one finishes the
>> next goes, so I don't think it's a bandwidth issue on the bus. The
>> files are on another server shared by nfs, on 100MBs FD switched
>> network.
>>
>> Justin
>
>
> --
> Darin Perusich
> Unix Systems Administrator
> Cognigen Corp.
> darinper at cognigencorp.com
>
>
>
> --
> This message has been scanned for viruses and
> dangerous content by Dynabrade using Mailscanner,
> and is believed to be clean.


-- 
-------------------------------------------
Justin Bennett
Red Hat (Linux) Certified Engineer
Network Administrator
Dynabrade Inc.
8989 Sheridan Dr
Clarence, NY 14031
716-631-0100 ext 215




More information about the nflug mailing list