New SGI disk mounting scheme; 1/3/03


Greetings,

    This message is to inform users of the SGI NIS cluster that the disk
    mounting procedure that we are using has now changed (hopefully for the
    better).

    Previously, mounting of disks from one sgi onto another was
    accomplished when the /etc/fstab file (which defines both the local and
    remote disks to be mounted) was read at boottime.  The problem with
    this procedure is that sometimes other computers having disks to be
    mounted are not up and running.  This causes the mount to fail, which
    as many of you know, can cause trouble (in terms of not being able to
    access one's files).  The only way to fix the failed mount under this
    scenario is to bring up the computer whose disk is to be mounted, and
    then manually issue a mount command as root (which is a pain for me).

    The new mounting system is called autofs.  This procedure is nice in
    that the mounts are dynamic, where mounting and unmounting occurs
    automatically as needed.  Thus, under autofs, if a user is logged onto
    instinct, and then types cd /inept2, this will automatically initiate a
    mount of the /inept2 disk.  Note that under this scenario, all that is
    required is that inept is up and running at the time the user types the
    cd command (clearly if inept is not up and running, the cd command will
    still fail).

    One detail is that you should be aware of is how the mounts are made.
    Under autofs, all mounts are made under
    /hosts/name_of_host_sharing_the_disk/disk_or_directory_to_be_mounted
    where name_of_host_sharing_the_disk is the name of the host sharing the
    disk (like instinct, inept, bioc01, nmrfac) and
    disk_or_directory_to_be_mounted is the name of the disk to be mounted
    (like instinct1, inept2, bioc011, nmrfac1).  As an example, if user
    ahinck logs into the machine inept, you would find that the result of
    typing the pwd
    command is

    /hosts/instinct/u/people/ahinck

    rather than /u/people/ahinck as before.  NOTE, HOWEVER, it is not
    necessary to specify the
    /hosts/name_of_host_sharing_the_disk when specifying a disk location.
    Thus, if user ahinck was logged onto inept, and he wanted to change
    back to his home directory, he could simply type

    cd /u/people/ahinck (cd alone would work as well)

    rather than the more cumbersome /hosts/instinct/u/people/ahinck.

    The reason for this is that symbolic links have been added to enable
    the simpler designation.  Thus, on the computer inept, a symbolic link
    (called /u) has been added at the root level (/) which points to the
    directory
    /hosts/instinct/u.  Thus, typing cd /u is equivalent to typing cd
    /hosts/instinct/u.  The same is true for all of the other shared disks.
      [[ It is possible that I have made a mistake in typing in one of the
    symbolic links, so if you think you find something that does not appear
    to be working as expected, please let me know and I will check it out
    ]].

    In summary, the new system should be much more reliable than the old
    without any real practical differences compared to the old system
    (owing to the symobolic links, the disk navigation is identical to that
    used under the old system; the only thing different you should notice
    is a different result from the pwd command]]

    Please, note that further steps are being taken to improve the
    reliability of the SGI NIS cluster.  Soon all of the SGI NIS computers
    (except for nmrfac) will have be supplied with emergency electrical
    power.  Since all of the SGIs (except nmrfac) also have UPSs, this
    means that in theory, none of the computers should ever shutdown in the
    case of an electrical outage (either shortterm, or longterm).  I
    anticipate that the switchover to emergency power will be accomplished
    without any interruption in service to the cluster; however if this
    changes, I will be sure to let you know.

    Andy