The vinum softraid driver should now be operational with devfs
    in the master repo.  vinum should now be able to handle full 64 bit
    geometries (multi-terrabyte disk subsystems).

    In addition, vinum configurations may reference devices by serial
    number or devtab label.  for example:

        drive drive0 device raid0.s1d
        drive drive1 device raid1.s1d
        drive drive2 device raid2.s1d
        drive drive3 device raid3.s1d
        volume myvol
            plex org raid5 128k
                sd drive drive0 length 50g
                sd drive drive1 length 50g
                sd drive drive2 length 50g
                sd drive drive3 length 50g

    And in /etc/devtab:

        raid0           serno   WD-WCAVY0248702
        raid1           serno   WD-WCAVY0250739
        raid2           serno   WD-WCAVY0245831
        raid3           serno   WD-WCAVY0246122

    The vinum configuration stored on-media stores translated values,
    so on-media the devices will be stored by serial number.  e.g.
    /dev/serno/WD-WCAVY0246122.

    vinum roots may or may not still be broken, only because we do not
    have a getsynthvnode() function which can find a drive by its serial
    number.

    Generally speaking, anyone using vinum will want to reference devices
    by serial number now to avoid renumbering problems.

    TASKS STILL TO DO (any takers?):

    * Add a feature to getsynthvnode() to locate a drive by its serial
      numbers.

    * Add general disk_create() support for vinum volumes.  vinum
      historically insisted on being the partition manager for filesystems,
      but that isn't always convenient.

    * RAID-5 issues.  vinum isn't smart about RAID-5.  Have you every
      tried to rebuild parity on a 10TB vinum drive?  We may need a
      SOC project to fix this... that is, sub-divide the space into
      smaller chunks and keep track of activity in the chunks so a rebuild
      does not have to rebuild the entire freaking topology.

                                                -Matt

Reply via email to