How to use smartctl with cciss

[Richard Ray]

I have a ML530 G2 with a 5300 array controller.
There are 6 drive connected to each channel.
When I run "smartctl -i -d cciss,0 /dev/cciss/c1d0" I get

smartctl version 5.37 [i686-pc-linux-gnu] Copyright (C) 2002-6 Bruce Allen
Home page is http://smartmontools.sourceforge.net/

Device: COMPAQ BD146863B3 Version: HPB6
Serial number: B8F74LTM
Device type: disk
Local Time is: Fri Jul 13 08:32:27 2007 CDT
Device supports SMART and is Enabled
Temperature Warning Enabled


When I run "smartctl -i -d cciss,0 /dev/cciss/c1d1" I get

smartctl version 5.37 [i686-pc-linux-gnu] Copyright (C) 2002-6 Bruce Allen
Home page is http://smartmontools.sourceforge.net/

Device: COMPAQ BD146863B3 Version: HPB6
Serial number: B8F74LTM
Device type: disk
Local Time is: Fri Jul 13 08:32:57 2007 CDT
Device supports SMART and is Enabled
Temperature Warning Enabled


If I query a channel 0 or channel 1 device I get information for the same drive.
This occurs for all drive pairs.
I am running Fedora Core 5 kernel 2.6.18.
Any suggestion?

[Stuart Browne]

My guess would be that the RAID controller has abstracted the individual disks, so it doesn't matter which logical disk you try to get from the controller, it see's the count the same.

Try:

smartctl -i -d cciss,1 /dev/cciss/c1d0

The ',1' after the disk 'type' is supposed to reference the physical disk on the controller, as against the logical disk number of the controller.

Also, even though you're using FC5, see if you can get the 'hpacucli' working on it. It gives much better information for virtual disks etc..

[skt_skt]

Acnt that be same physical drive connected through diffrent channel/controller/paths

[Sac_3]

Hi Richard,

Pls check the below URL. Hope this helps you:

https://packages.debian.org/unstable/cpqarrayd

Regards,
SaC

P.S: Best way to thank in this forum is to assign points

[Van den Broeck Tijl]

You're using it in the right way. Using smartctl you can monitor a physical disk behind the controller. As Santhosh noticed, depending upon your configuration you could be monitoring the same disk twice trough a different logical path.

Anyhow, the c0d0-part is the representation of your logical disks, the physicals behind them are queried in the "cciss,0" part, where the phys. disk no. is specified as "0" ... or 1 in cciss,1. To configure monitoring in smartmontools, add the disks you'd want to monitor in /etc/smartd.conf

# Comment out DEVICESCAN, it's a BAD idea for controller based systems.
/dev/cciss/c0d0 -f -M daily -m yourmail@somehost.com -d cciss,0 -a -s L/../../7/04
/dev/cciss/c0d0 -f -M daily -m yourmail@somehost.com -d cciss,1 -a -s L/../../7/04

This will monitor disk 0 & 1 of my logical Mirror disk on the controller. Automatic testing each sunday night and mail error events to the speficied e-mail.

Imho smartmontools is a better idea than the HP-tools disk monitoring as they tend to be somewhat resource consuming. A pretty good alternative is cpqarrayd which you can find here: http://www.strocamp.net/opensource/
It monitors the cciss event & device log and logs it to your syslog and sends SNMP traps.

[Richard Ray]

If I run "smartctl -i -d cciss,1 /dev/cciss/c1d0" and "smartctl -i -d cciss,1 /dev/cciss/c1d1" I get the same info.

If I run "smartctl -i -d cciss,2 /dev/cciss/c1d0" and "smartctl -i -d cciss,2 /dev/cciss/c1d1" I get the same info.

It's the same for all drive pairs.

I installed 'hpacucli' and it works but I don't see a way to have the drive perform test as in "smartctl -t long -d cciss,0 /dev/cciss/c1d0"

Scheduling regular test in /etc/smartd.conf such as " /dev/cciss/c1d0 -f -M daily -m yourmail@somehost.com -d cciss,0 -a -s L/../../7/04" and /dev/cciss/c1d1 -f -M daily -m yourmail@somehost.com -d cciss,0 -a -s L/../../7/04" results in the same drive being scanned twice.

If I run "smartctl -t long -d cciss,0 /dev/cciss/c1d0" and then "smartctl -l selftest -d cciss,0 dev/cciss/c1d0" I see "# 1 Background long Self test in progress ... - NOW"
If run "smartctl -l selftest -d cciss,0 dev/cciss/c1d1" I see "# 1 Background long Self test in progress ... - NOW"

[David Claypool]

My guess is that 'smartctl' doesn't know how to properly talk through a hardware RAID controller. Additionally, you are referencing the logical drives with your /dev.

/dev/cciss/c1d0 is the first logical drive on RAID set c1.

/dev/cciss/c1d1 is the second logical drive on RAID set c1.

[Van den Broeck Tijl]

@David, actually it does support that. Supported it for some time on 3Ware's and since 5.36 for SmartArray's as well. You're definitly right about the logical drives.

Richard, can you give us a layout of the logical & physical drives and their organisation and what smartctl tells you? If you're using identical physical disks you won't notice any "difference" as smartmontools is focused on monitoring physical disks and not controllers or logical disks (you can and will not find any information of those in smartmontools). If you want that as well, cpqarrayd is really nice.