StoreEver Tape Storage
1823015 Members
3765 Online
109645 Solutions
New Discussion юеВ

Re: Problem using LTO-1 on Linux

 
OlivierVigneresse
New Member

Problem using LTO-1 on Linux

Hi,
I have a DL380-G3 server with an Ultrium 230 external LTO drive attached to it; the server runs Ubuntu LTS v6.0
Whilst the LTO drive works, I can only get about 50GB on each tape - a far cry from the 100-200 that I'd expect to get.

The tests/diags I've carried out so far:
- I had to configure the cciss driver as the external drive is attached to the built-in RAID controller in the DL380
- I use bacula as a backup software, and have carried out various tests with tar as well: in both cases, the tape is full after about 50-55GB
- I have carried the tests with the / partition, and also a /vm partition which contains a single 60GB file = same result = so I know that it i not some artifact because of file system overhead or anything like that
- I can successfully write and read back, so the drive and tapes seem OK - I've tried 3 different (new) tapes - same thing - so I don't think the problem is hardware related.

I Google'd a bit and found a page with non-standard densities for LTO drives - when running a "mt -f /dev/nst0 status" I had by default a density code of 0x24 - I changed that to 0x40 (suitable for LTO-1 rives I believe) - exactly the same thing happens: tape is full after 50-odd GB

I have a sneaky feeling that the problem is related to density or other configuration which causes the st driver to think that the tape is full way before reaching the end, but I just cannot figure out a way forward...

If anyone has experienced this before, or has any suggestion, the help will be greatly received - I updated from a DLT-VI drive when I got the DL380/LTO-1, and so far there is hardly any difference :-(

Regards,

Olivier
5 REPLIES 5
Lewis Finch
Honored Contributor

Re: Problem using LTO-1 on Linux

I suspect you have a lot of "shoe-shining" going on. The tape starts then stops then starts again. This results in lots of blank space on the tape. To start with, take it off the RAID controller and get it on it's own separate SCSI controller. RAID controllers don't play nice with tape drives.
"You can't lead the orchestra without turning your back to the crowd"
OlivierVigneresse
New Member

Re: Problem using LTO-1 on Linux

Thanks for the reply, Lewis.
The DL-380 comes only with an RAID controller, and I have a separate SCSI controller lying around, but it is also a RAID one (and old Compaq one) - is there a way to configure a RAID controller to avoid shoe-shining? And if not, is there any particular SCSI-but-non-RAID controller that I should try to get hold of?
Thanks again for your help
Lewis Finch
Honored Contributor

Re: Problem using LTO-1 on Linux

This one should work just fine

http://www.hp.com/products1/storage/compatibility/tapebackup/ThirdParty/index.html#HBA
"You can't lead the orchestra without turning your back to the crowd"
OlivierVigneresse
New Member

Re: Problem using LTO-1 on Linux

Thanks again.
I'll try to find one in the next week or two and will keep you posted of my progress!
Regards,
OlivierVigneresse
New Member

Re: Problem using LTO-1 on Linux

Finally got hold of a non-RAID adapter - namely an Adaptec 29160... in a nutshell, nothing's changed.

I have also connected to the 29160 my previous (DLT) drive, and get 15GB out of a IIIXT tape - bang on what's expected in uncompressed mode (I used dd if=/dev/urandom of=/dev/nst? for the test, so did not expect to get any compression).

But when using the Ultrium-230 drive, I get still only 50GB on the tape before it's seen as full (tried twice, with brand new tapes - so I'm reasonably sure that it's not a faulty-tape issue).
Not sure whether that's relevant, but running mt -f /dev/nst0 status I get:
Tape block size 0 bytes. Density code 0x40 (unknown).
and an mt -f /dev/nst0 densities returns a list where 0x40 is not defined - and there's no other code for LTO/Ultrium tapes.

With the DLT working as expected, I have a feeling that the hardware side is OK and that it is the Linux side (probably something having to do with either the st driver or the SCSI layer as a whole) that causes the problem... but I just cannot figure out where to start.

FYI, the server is running Xubuntu (Ubuntu with XFCE), with a 2.6.15-26 kernel; other version numbers are:
mt (GNU cpio 2.6)
dd (coreutils) 5.93

Has any body experienced a similar problem, or have an idea for a fix?

Many thanks,

Olivier