On Wed, Jan 7, 2009 at 17:12, Volker A. Brandt <v...@bb-c.de> wrote: >> The Samsung HD103UJ drives are nice, if you're not using >> NVidia controllers - there's a bug in either the drives or the >> controllers that makes them drop drives fairly frequently. > > Do you happen to have more details about this problem? Or some > pointers? We have 3 x2200m2 servers that we added pairs of these drives (specifically, the HD753UJ variant: 750GB instead of 1TB) to. We set up small (40G or so, I forget; we didn't really need the space, but buying smaller disks wasn't significantly cheaper) SVM mirrors on two of these machines, and a small SVM mirror plus a large zpool on the third. Within two weeks, all three machines had dropped a disk in some manner. The behavior we saw goes like this: metastat reports errors, output of 'format' changes for the dropped disk but still shows the disk. If the disk is moved to another machine (a different chipset; i.e., with another controller) then it shows up fine, all data intact, everything hunky-dory. We didn't lose data, but we did lose an SVM array and had to restore from backups.
We replaced the drives with 4 Maxtors and 2 Seagate ES2s. None have reported problems yet. I don't know of any other solution, if you don't want to add a controller. It doesn't appear to be a problem with the drives, or a problem with the chipset, but the combination of drive+chipset causes wonkiness. Google shows some users having problems with this under XP, so it's probably not just a driver issue. This was what made me suspect the combination was a bad one, and further testing shows that that's probably the case: the drives work on other controllers, and other drives work on these controllers. The drives themselves are still working fine; we moved them to a SCSI->sata jbod with a non-nV controller and they're happy there. Will _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
