Hi, Raymond wrote on Wed, Jun 07, 2006 at 11:39:34AM +0200: > 2006/6/7, Stuart Henderson <[EMAIL PROTECTED]>: >> On 2006/06/06 17:48, Mike Spenard wrote:
>>> I have a Dell Poweredge 850 >>> with the Dell CERC SATA 1.5/6ch controller. >> It's probably an aac(4), for which you'll need a custom >> kernel - it was taken out of GENERIC for a reason. Concluding from http://developer.novell.com/yes/71337.htm this is actually an Adaptec 2610SA, i.e. in fact aac(4). >> If you're unable to replace it with, e.g. some lsi megaraid >> sata card, you'll have to install an old OS and transfer across >> a newer release and suitable custom kernel from another box. Don't do that unless reliability is not required. >> Don't be entirely surprised if it wedges occasionally, this is >> not a recommended controller. > I made my own release with an aac enabled kernel. Don't use that. Read the archives. * The controller firmware is buggy. It crashes randomly, the frequency depending on the firmware version in use, but hardly or not at all on load. With the firmware version 7348 (Dec. 2004), typical crash frequencies are once in a few weeks, with firmware version 7244 (May 2004), up to once in a few hours. I did not test the newest version 8205 (Aug. 2005). See http://www.adaptec.com/en-US/support/raid/ --- click --> controller type --- click --> bios updates --- click --> firmware version 8205 WHEN INSTALLING ADAPTEC FIRMWARE UPDATES, KEEP BACKUPS OF BOTH THE OLD FIRMWARE AND YOUR DATA. I did once encounter an Adaptec firmware update performing much worse than the previous one. I did once encounter an Adaptec firmware update trashing the logical volume setup of the array (including the file systems and data, of course). FIRMWARE UPDATES FOR THE 2610SA ARE NOT AVAILABLE FROM ADAPTEC. I do not know whether it makes sense to try the 2410SA or 2810SA updates. I do not know whether Adaptec firmware updates work on DELL rebranded devices. MAKE SURE ANY CONTRACTS WITH DELL YOU HAPPEN TO HAVE SIGNED DO NOT FORBID YOU TO UPGRADE FIRMWARE. * When the controller firmware crashes, the operating system must wait two minutes (= 120 seconds) for the controller firmware to reboot and reinitialize. During that time, no SCSI access to the contoller is possible, which means you cannot access your data while waiting. When i last checked, OpenBSD aac(4) only waited one minute (= 60 seconds). This resulted in completely freezing the kernel. You did not even get a kernel panic. I don't know whether the work done by nate@ in Nov. 2005, see http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/ic/aac.c & http://www.openbsd.org/cgi-bin/cvsweb/src/sys/dev/pci/aac_pci.c changed this behaviour. I don't claim this is either easy or difficult to fix. I don't claim it this is the only issue. But i do suspect all this is somewhat like groping for black cats in the dark cellar without knowing how many there are. > I'll send you the link. I believe it was taken out because the > driver wasn't going to be good enough without proper documentation. This is not wrong, but you are flattering these Adaptec devices more than they deserve. > So use it at your own risk. Rather, do not use it unless reliability is not an issue for your purpose. Hope that helps, Ingo -- Ingo Schwarze <[EMAIL PROTECTED]> usta.de / studis.de sysop
