• Main INDEX
  • Monthly INDEX
  • PREV
  • NEXT
    Make New Entry, Make Followup Entry

    User name R. Michaels

    Log entry time 17:43:56 on November 2,2010

    Entry number 338201

    This entry is a followup to: 338184

    keyword=re: disk full problem and amelioration.

    I looked at the data file handling and found nothing really wrong,
    but the following "perfect storm" occured

    1. Yesterday someone rebooted adaql2. The MSS copying had registered
    in a log file that it was working, so when another instance of the
    script started it exits and sends me a mail. This is normal
    (prevents 2 instances of the script) but I was too lazy to reset
    the log file until 11 am today, so there was a period of time
    when we were not copying to MSS and the disks got dangerously full.
    (Hmmm... I should clear the log file when the computer reboots.)

    2. DVCS is taking a *HUGE* rate of bytes/sec.

    -------------- Solution -------------------------

    What I did:

    1. Copying script launched twice as often, should go faster.

    2. Deletion is more agressive. The percentage of available disk
    space governs how long a file may stay on adaql2 disks, and the
    criteria is more strict now, i.e. for a given percentage the
    time allowed is a lot smaller. The time is the time of last
    usage, so it should not delete files being analyzed.

    ---------------- Comments -------------------------

    Some files had been moved to a work disk /adaql4/work1/adaql2.
    Do not delete these. I'll make sure they exist in MSS. Never
    delete files from adaql2. Instead, if desperate, "move" them
    to a work disk.

    BTW, I got paged (and e-mailed) automatically when the disks filled,
    so at least something worked.