Date: Fri, 18 Nov 2011 13:11:22 -0800 (PST) From: Doug Ambrisko <ambrisko@ambrisko.com> To: =?UTF-8?Q?Dag-Erling_Sm=C3=B8rgrav?= <des@des.no> Cc: scsi@freebsd.org Subject: Re: smartd blows away mfi config Message-ID: <201111182111.pAILBMmq059649@ambrisko.com> In-Reply-To: <86d3cqpxh0.fsf@ds4.des.no>
next in thread | previous in thread | raw e-mail | index | archive | help
Dag-Erling Sm?rgrav writes: | I have a Dell PowerEdge 1950 II with a PERC/5i controller and two SATA | disks. I had no trouble at all setting up logical volumes with mfiutil | and installing FreeBSD 8.2, but when smartd starts (with smartd.conf | consisting only of "DEVICESCAN") it seems to blow away the config so the | virtual disks disappear from underneath the running system and I have to | powercycle the machine. When the machine reboots, the controller does | not find any logical disks. The configuration utility finds a "foreign | configuration" which I can import, and there doesn't seem to be any data | missing. | | This does not seem to happen if I specify the passX devices explicitly | in smartd.conf - only if I use DEVICESCAN. This indicates that the | problem arises when smartd scans the SES device (i.e. the backplane). | The problem is that the unit numbers can vary over time, e.g. if I add | an optical drive or boot the system with a USB mass storage device | attached, so listing them explicitly is not a good long-term solution. | | Any suggestions? The ideal solution would be a patch for either the | kernel or smartd to prevent it from blowing away the mfi config, but | short of that, is there a way to have smartd ignore the backplane when | scanning for supported devices? A drivedb entry, perhaps? As a last | resort, I guess I could regnerate smartd.conf at boot time by parsing | dmesg and / or the cam device list. I suggest that as a test bump "up" the log level of the RAID controller via: hw.mfi.event_class="-2" in loader.conf or at the the loader prompt. This will give you real-time knowledge of what the controller is thinking. I notice that the firmware tends to see lots of sense issues when going behind it's back and it sometimes doesn't like that. You can also use MegaCli to dump the controller log but just won't see the messages in real time. They should have a time stamp on them. My guess is that the during the enclosure probe from smartd, the firmware sees the enclosure go away and then come back making it think it is a foreign set of disks. This could help isolate what the issue is. It isn't a solution to the problem. Doug A.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201111182111.pAILBMmq059649>