Those certainly look OK. I've read over my admin notes from earlier this year. Unfortunately not very good notes. I got most things working, but it's not clear I really knew what I was doing :-( I'm still unable to get sd.conf to set the physical size for my 3 TB USB disk correctly despite making certain that the strings *exactly* match what the source code wants.
What OS release was the pool built with? Could it be that an update has resulted in zfs not being willing to do what it did when the pool was created? In that case rolling back to the BE that built the pool might let you add the disk in w/ ashift=9. However, I don't think that changes the need to rebuild the pool w/ ashift=12. For ashift=9 to be correct prior to the firmware update, the controller would have to have been doing the RMW buffering internally. It seems strange that they would remove such a feature. You might want to look for messages in old log files from prior to the update if you still have them. My notes don't reflect it, but I seem to recall that I managed to build an improperly aligned pool, but did not see any console messages and was not looking at /var/adm/messages until performance testing showed very poor performance. It may be you've stumbled across an old problem you didn't know you had. Good luck, Reg -------------------------------------------- On Thu, 11/21/13, Francis Swasey <[email protected]> wrote: Subject: Re: [OpenIndiana-discuss] zpool replace says the disk has a different sector alignment To: "Discussion list for OpenIndiana" <[email protected]> Date: Thursday, November 21, 2013, 6:20 AM On Nov 20, 2013, at 3:26 PM, Reginald Beardsley <[email protected]> wrote: > > What does: > > echo ::sd_state | mdb -k | egrep '(^un|_blocksize)' > > report? un 1: ffffff08ff561300 un_sys_blocksize = 0x200 un_tgt_blocksize = 0x200 un_phy_blocksize = 0x1000 un_f_tgt_blocksize_is_valid = 0x1 un 2: ffffff091e10c080 un_sys_blocksize = 0x200 un_tgt_blocksize = 0x200 un_phy_blocksize = 0x1000 un_f_tgt_blocksize_is_valid = 0x1 un 3: ffffff090dfea300 un_sys_blocksize = 0x200 un_tgt_blocksize = 0x200 un_phy_blocksize = 0x1000 un_f_tgt_blocksize_is_valid = 0x1 (that's the first three disks in the jbod. "un 1" is the new disk. > > In particular does the kernel think the disk parameters for old and new drives are the same? Yes, it does. Frank -----Inline Attachment Follows----- _______________________________________________ OpenIndiana-discuss mailing list [email protected] http://openindiana.org/mailman/listinfo/openindiana-discuss _______________________________________________ OpenIndiana-discuss mailing list [email protected] http://openindiana.org/mailman/listinfo/openindiana-discuss
