Just Lost an Hour and a Half of Work

Sitting at work.  Forgot the cardinal rule of doing work on the share drive, which is DON’T DO WORK ON THE SHARE DRIVE.

This has been a public service message from a guy who just lost an hour and a half of work, working on the share drive.

Gah.

2+
avataravatar

8 thoughts on “Just Lost an Hour and a Half of Work”

  1. 10 Cents:
    Thanks for sharing this with us.

    What happened? Or do I have to buy the book?

    The text describing this incident is what I lost.  Now it never happened.

    2+
    avataravatar
  2. I presume this is some kind of Microsoft thing.

    Ever since May 2000 I’ve kept almost all of my project and archival files on an NFS shared partition on the current in-house server machine.  (Over the 18 years, this has migrated across four machines (Moby, Ceres, Pallas, and Juno), two vendors (Sun and Dell), and three CPU architectures (SPARC, IA-32, and IA-64).  The big exported partition is a RAID array, currently RAID-6 with 8 SSD drives which can run with two drives failed.  This is mounted read/write on client machines as “/files” and looks like this from the client:

        $ df -h /files
        Filesystem      Size  Used Avail Use% Mounted on
        juno:/home       11T  1.6T  8.9T  15% /juno
    

    The backup tape drive is on the same server as the shared file system, which makes backups much faster and doesn’t clog the network with data traffic while large backups are running.

    This allows the client machines to be very lightweight in their file system footprint, basically just the operating system and utilities install, and provides identical access to the same /files environment regardless of which client you’re using.  Sure, it isn’t as fast as accessing data from a local drive, but the things I tend to do are usually compute-intense rather than big data.  When I do something with enormous files (for example, reducing NASA planetary imagery for use with Earth and Moon Viewer/Solar System Explorer), I just run the jobs on the in-house server itself, where they have local access to /files.

    The NFS protocol dates to 1984, with updates over the years.  I have found it to be as robust as anything involving networks and multi-vendor support to be.  As long as you don’t use fancy new features but simply use it as a shared file system, it is, in my experience, utterly bulletproof.  The last major innovations I use are NFS over TCP and 64-bit file addresses, which were introduced in NFSv3 in 1995.

    6+
    avataravataravataravataravataravatar
  3. I used to use rsync to keep podcast subscriptions and playlists up to date on a hacked iPod Nano (tall square one) running RockBox firmware.  That way I could do all the management and downloading on the home computer (of course) and just sync the thing in the morning before doing the motorcycle+train+walk commute cycle to Tokyo.

    It was during this time that I implemented a SimpleTron in Chipmunk Basic on the Palm from memory — in my spare time.

    0

Leave a Reply