Mine says:
st1: Block limits 1 - 16777215 bytes.
But in the process of finding that, I may have found the problem. I saw
some "st1: Write not multiple of tape block size." messages in the dmesg
output. So I checked on the drive's status:
# mt -f /dev/nst1 status
SCSI 2 tape drive:
File number=14, block number=0, partition=0.
Tape block size 0 bytes. Density code 0x46 (no translation).
Soft error count since last status=0
General status bits on (81010000):
EOF ONLINE IM_REP_EN
So, setblk was set on 0 (variable block size) by default. In this
condition, dd/tar block size 512k yields 63MB/sec and all attempts to
use a block size bigger than 512k with tar or dd fail.
If I do:
mt -f /dev/st1 setblk 1k
Results: block sizes for dd and tar all seem to work (tested up to
4096k) but transfer speed is at 16MB.
mt -f /dev/st1 setblk 256k
Results: block sizes for dd and tar all seem to work (tested up to
4096k) and transfer speed is 119MB.
mt -f /dev/st1 setblk 512k
Results: block sizes for dd and tar all seem to work (tested up to
4096k) and transfer speed is 105MB.
It seems like when it is set to 0 (variable block size) it can't handle
anything bigger than 512k from the software layer (dd or tar)
gracefully. Further testing yields that 512k may be the hardware block
size limit of the drive, as trying to feed anything higher to setblk
yields the "device or resource busy" errors.
So it seems like I am on my way. Thanks for the suggestion which
ultimately led me to discover this. :-)
-Fran
Fran Fabrizio wrote:
This isn't strictly related to Amanda but this list has been an
excellent resource for tape drive configuration in the past, so I am
hoping someone here can help.
I've recently obtained an LTO4 drive to upgrade my Amanda backup
infrastructure. I've installed it into our changer and all the basic
tests check out, and now I am trying to adjust the block size for
performance. I'm testing options by writing 10GB files to the tape
using dd and tar with various block sizes. It seems I am hitting a
limit of 512k on my block size - any more and I get "device or
resource busy" errors from the dd and tar commands.
I am thinking I might have some sort of OS limitation. My OS is
CentOS 5 and my kernel on the tape server at the moment is
2.6.18-53.el5PAE. Does this sound familiar to anyone?
Any other causes of this? I was wondering if this was somehow a
parameter of the tape drive but using block sizes of 1MB and 2MB seems
very common for LTO-4 tape drives.
Appreciate any thoughts you all might have. It's working perfectly
with 512k blocks, but performance is at about 63MB/sec and I think I
should be able to get some more out of the drive.
Thanks,
Fran
--
Fran Fabrizio, Senior Systems Analyst
Department of Computer and Information Sciences
University of Alabama at Birmingham
http://www.cis.uab.edu/ | 205.934.0653
