Description of problem: SMART monitoring (smartd) for disks connected to HP/Compaq SmartArray controllers doesn't seem to work correctly. It seems to be monitoring only the last disk specified in the configuration file. For example, I have the following in my "/etc/smartd.conf": /dev/cciss/c0d0 -a -d cciss,0 -s (S/../.././01|L/../../7/06) -m root /dev/cciss/c0d0 -a -d cciss,1 -s (S/../.././02|L/../../7/05) -m root /dev/cciss/c0d0 -a -d cciss,2 -s (S/../.././03|L/../../7/04) -m root /dev/cciss/c0d0 -a -d cciss,3 -s (S/../.././04|L/../../7/03) -m root /dev/cciss/c0d0 -a -d cciss,4 -s (S/../.././05|L/../../7/02) -m root /dev/cciss/c0d0 -a -d cciss,5 -s (S/../.././06|L/../../7/01) -m root There are self-tests configured for all disks, but smartd is running them all on "cciss,5", as can be seen on this disk's selftest log. I've seen this on two different machines (one i386, another x86-64), each one with different controller models (one SCSI, another SAS, both of them "cciss" of course). It is important to notice that "smartctl" seems to do the right thing. It's only "smartd" that doesn't work.
Just a small note: I've confirmed that it isn't just the self-test functionality that's broken on "cciss". "smartd" only monitors the last disk, which can be seen from the messages log, where smartd reports changes in disk temperature for all disks, with the exact same value (where "smartctl" shows that the disks really have different values for this attribute).
I'm trying to reproduce the bug, but I get this message in the log for all the disks: Device: /dev/cciss/c0d0 [cciss_disk_00], does not support SMART Self-Test Log. I'm using smartmontools-5.36-3.1.el5 on a machine equipped with Compaq Computer Corporation Smart Array 5i/532. Could you please provide the part of /var/log/messages with smartd startup log? What version of smartmontools do you use?
Created attachment 239161 [details] smartd startup log with an HP SmartArray 5i/532 controller
The attached file (smart.log) has the information you requested, taken from a machine with an 5i/532 controller also. I'm also using smartmontools-5.36-3.1.el5, sorry for not mentioning it before. Maybe _only_ your last disk (in smartd.conf) doesn't support SMART Self-Test Log, in which case you are already reproducing the bug. :) Since smartctl works correctly, it can be used to prove or disprove that. I've build the SRPM from Fedora 7 updates (smartmontools-5.37-3.2.fc7) and it works correctly. However, it would be nice to have this patched in RHEL proper, since these controllers are so popular in datacenters (and some admins, like myself, aren't very keen on installing all sorts of crap from hardware vendors just to check if their disks are developing bad sectors...).
No. Even smartcl says "Device does not support Self Test logging" for every disk. I'll try to find some other testing machine but this doesn't look too promising.
cciss support has been added in rhel 5.3 update