On Sep 13, 2013, at 11:46 PM, Hugh McIntyre <li...@mcintyreweb.com> wrote:
> I have a related question to the recent discussion of ashift=12 and advanced > format disks. The problem is that I have 4 disks attached via a LSI 9211-4i > SATA/SAS adapter (which are fine) but several others attached via motherboard > SATA connections and running in IDE compatibility mode. > > The problem is that in IDE mode, the disks don't report any geometry > information via format "inq", or "iostat -Er | grep -i vendor", and I only > get the LSI-attached disks (plus a motherboard attached DVD drive) listed > from "echo ::sd_state | mdb -k | egrep '(^un|_blocksize)'". > > I'm sure all of these symptoms are linked, i.e. if the kernel does not know > the disk geometry then iostat and format are not going to report it either. > But this seems like a problem for ashift detection logic that is not going to > be able to get the geometry either. > > (As an aside, no doubt I should have switched from IDE mode long ago ... but > this is the default :( ) > > So my two questions: > > 1. If I change the SATA ports to AHCI instead of IDE in the BIOS, I'm > guessing this will fix the bug, but I also assume this will change the paths > under /devices and therefore make the system unbootable because one of these > disks is a SSD containing the rootfs? Is there a safe way to do this, or can > I assume the paths will stay the same? And would I get better performance in > AHCI mode nowadays? (This is a network-limited backup box so performance is > not critical, but one partition of the SSD contains L2ARC used for dedup of > some of the data). Do this. You won't notice performance improvements on rootfs, not because they aren't there, but because root rarely impacts system performance anyway. To make the boot environment happy, you might have to: 0. change BIOS setting to AHCI 1. boot livecd with shell login 2. import rpool 3. export rpool 4. boot This procedure has been documented hundreds of times in the ZFS archives -- richard > > 2. Is this a bug I should file, with the possbility of being fixed, or just a > limitation of the IDE layer that will never be able to get disk geometries > and detect ashift cleanly? > > Thanks, > > Hugh. > > PS: in my case, I worked around this by moving the disks around long enough > to get the geometry, but a long-term answer would be good. > > >> https://www.illumos.org/issues/2665#note-3 >>> http://wiki.illumos.org/display/illumos/ZFS+and+Advanced+Format+disks > > _______________________________________________ > OmniOS-discuss mailing list > OmniOS-discuss@lists.omniti.com > http://lists.omniti.com/mailman/listinfo/omnios-discuss -- richard.ell...@richardelling.com +1-760-896-4422
_______________________________________________ OmniOS-discuss mailing list OmniOS-discuss@lists.omniti.com http://lists.omniti.com/mailman/listinfo/omnios-discuss
