On Wed, 26 Oct 2005, Arnaud Quette wrote:
> About your last question ("Are there immediate-poweroff UPSes that don't
> ..."), I don't know for the moment. The RAID problem is known and
> documented. For the drives, these should already be flushed as we're
> remounted RO...That is different from a drive cache flush command, and component devices of MD devices won't be flushed without it, since those are NOT clean and read-only at S99. Even if you sent the proper flush commands to the devices that support or need it (and there is the PATA, SATA and SCSI variations of those...), it'd still screw up MD RAID arrays. If you want to support dumb UPSes properly, you have to do so in the kernel as the last thing before a ACPI/APM shutdown or final halt. IMO, the best way to support poweroff with dumb UPSes is to not power them off at all. Let them run near dry and power off the load by themselves. -- "One disk to rule them all, One disk to find them. One disk to bring them all and in the darkness grind them. In the Land of Redmond where the shadows lie." -- The Silicon Valley Tarot Henrique Holschuh -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
