Ultrium 1760 SAS on linux

[ipv4]

Having trouble changing media

I have an Ultrium 1760 SAS tape drive connected to a HP SmartArray P800 controller -- inside a DL380 G6.

I have put the following in my /etc/rc.local script to initialize the drive on startup.

# modprobe osst
# echo "engage scsi" > /proc/driver/cciss/cciss1

I am able to work with the drive using mt and tar commands at this point. Whenever I change tapes I am unable to use the drive anymore without doing a full reboot.

This is regardless of whether I do a "mt -f /dev/st0 offline", or eject, or just press the eject button.

What is the proper way to change media?

When I put a new tape in all commands fail like the following...

# mt -f /dev/st0 status
/dev/st0: Input/output error



And i see the following in dmesg:

[ 2950.990393] cciss1: resetting tape drive or medium changer.
[ 2953.520103] cciss1: a power on or device reset detected
[ 2953.525308] cciss1: a power on or device reset detected
[ 2953.530512] cciss1: device is ready.
[ 2953.534588] st0: Error 6 (driver bt 0x0, host bt 0x0).

[ipv4]

After switching tapes I have been able to get the drive functional again after performing the following.

# ./rescan-scsi-bus.sh --forceremove
# mt -f /dev/st0 retension
# mt -f /dev/st0 status

This workaround was found with a significant amount of trial and error. I still don't feel this is how the drive _should_ function.

After issueing the above commands i see the following in dmesg as the drive is coming back to life.

cciss1: resetting tape drive or medium changer.
cciss1: a power on or device reset detected
cciss1: a power on or device reset detected
cciss1: device is ready.
st0: Error 6 (driver bt 0x0, host bt 0x0).
st0: Block limits 1 - 16777215 bytes.


Anybody found a better way to switch tapes?

[AnthonySN]

I would use a separate scsi card rather than HP Smart Array P800 and leave it for raid funtionality alone.

[Rob Leadbeater]

Hi,

I would normally say the same, however as this is a SAS drive rather than SCSI, options are limited.

According to this page:

http://www.hp.com/products1/storage/compatibility/tapebackup/ISS/Detail/16-0046-0190.html#CONTROLLERS

The HP ProLiant DL380 G6, G7 does not have a suitable integrated controller available for connecting the HP StorageWorks Ultrium 1760 SAS.

Please use one of the following compatible adapters:

462828-B21 - HP Smart Array P212 Controller HP RECOMMENDED
462830-B21 - HP Smart Array P411 Controller (External only)
381513-B21 - HP Smart Array P800 Controller


What version of Linux are you running ?
It may be that an upgrade to the cciss driver would help...

Regards,

Rob

[JoeJoeDan · Accepted solution]

I have the exact same behaviour.

Proliant ML150G6, Smart Array Card P410 256MB with only the Ultrium connected to it. Ubuntu 10.10 64 bits

Most of the time, I can use the tape after doing:

echo "rescan" > /proc/scsi/cciss/4
mt -f /dev/st0 rewind
mt -f /dev/st0 status

I only reboot when the device is offlined, dmesg shows

...rejecting I/O to offline device.

[ipv4]

Rob - I have tried several vendor supplied kernels in the RHEL 5 range and the latest release from kernel.org compiled myself. Same behavior using every release.


I've very disappointed in the behavior I am seeing from this raid card and tape drive. I can't help but thinking there is a very simple workaround.

I'm wondering if commercial backup software would have the same behavior using this hardware combination. I find that a little hard to believe.

[JoeJoeDan]

If you're using Debian,
http://wiki.debian.org/HP/ProLiant describes a way to use the newer hpsa driver instead of cciss.

I have it loaded on my box and report back if it solves the problem or not.

[JoeJoeDan]

I've been using the new driver for two days now (backup, restore) and it's been working so far.

[Marino Meloni_1]

could this be related ?

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=bg&taskId=110&prodSeriesId=63986&prodTypeId=12169&objectID=c02236321

[ipv4]

I don't think that issue is related because I'm using an HP P800 controller which goes through the cciss driver to access the Tape drive as a SCSI device.

If i was using a SAS HBA that issue would be something to consider.

Thanks.

[Z37A]

I have the some problem with 2 of 4 servers. This two are using SLES11 SP1 and the other two are using SLES10 SP3 (in this work fine).

I'm using CA Arcserve Backup for Linux 11.5 SP3 in all servers.

Attach dmesg and tail from Brighstor log!

[ 230.033404] cciss1: resetting tape drive or medium changer.
[ 231.043988] cciss1: a power on or device reset detected
[ 231.043991] cciss1: a power on or device reset detected
[ 231.043993] cciss1: device is ready.
[ 310.848538] JBD: barrier-based sync failed on cciss/c0d0p1 - disabling barriers
[ 343.068880] cciss1: resetting tape drive or medium changer.
[ 344.085281] cciss1: a power on or device reset detected
[ 344.085284] cciss1: a power on or device reset detected
[ 344.085286] cciss1: device is ready.
SERVER:/dev/mapper # tail /opt/CA/BrightStorARCserve/logs/camediad.log
[12/07/2010-10:54:27 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0)] Proposed Media: TEST Seq: 1 ID: 3c92 serno:
[12/07/2010-10:54:39 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0)] Drv (4,0,0) Cmd=[SPACE] SCSI Error: unknown error.
[12/07/2010-10:54:40 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0)] Format successful.
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Main-8295] Job id 2 started.
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] ******************************************
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] CRBS: Begin checking block size via mode sense
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] CRBS: Current block size has been set to 0 by an unknown source. Will reset it back to 65536
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] CRBS: Mode Select successfully set the block size
[12/07/2010-10:56:31 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] ******************************************
[12/07/2010-10:56:33 ,-1,0,0,0,-1,2,3,0,0] [Grp -8867(GROUP0) JobID-2] Drv (4,0,0) Cmd=[SPACE] SCSI Error: unknown error.