none
Job history time RRS feed

  • Question

  • Hi,

    I added a significant amount of jobs to a HPC cluster, and later cancelled them.

    Now I cannot add more jobs because of the consistently appearing error message:
    "Could not create a new job because the database is full".

    Is there an easy way to delete old cancelled jobs?

    I have set the "Days to keep in Job History" setting to 1 day, but there are still some entries in the database from two months before. Also, some entries don't have a submission date at all, and don't get removed.

    Thanks for your help,
    Michael
    Wednesday, May 6, 2009 8:47 PM

Answers

  • I think you've hit a known issue where jobs that don't have an End Time never get cleaned up.  We should have this fixed in our forthcoming SP1.  In the meantime you can submit the jobs to let them start (and then cancel them).

    I  you need a better solution I'd recommend calling support; they can probably work with you to get the DB cleaned up (note that you won't have to pay if it's a product bug).

    As a last (or first?) resort, reinstalling the HPC Pack will clean out the DB and solve the problem.

    Thanks,
    Josh

    P.S. If you feel like spending some cash, you could also upgrade to full SQL which will have some performance benefits as well as having no size limit.
    -Josh
    Thursday, July 16, 2009 12:39 AM
    Moderator

All replies

  • if you create jobs without submitting them the will stay in the database. submit those jobs or cancel them.
    Sunday, July 12, 2009 12:37 AM
  • I think you've hit a known issue where jobs that don't have an End Time never get cleaned up.  We should have this fixed in our forthcoming SP1.  In the meantime you can submit the jobs to let them start (and then cancel them).

    I  you need a better solution I'd recommend calling support; they can probably work with you to get the DB cleaned up (note that you won't have to pay if it's a product bug).

    As a last (or first?) resort, reinstalling the HPC Pack will clean out the DB and solve the problem.

    Thanks,
    Josh

    P.S. If you feel like spending some cash, you could also upgrade to full SQL which will have some performance benefits as well as having no size limit.
    -Josh
    Thursday, July 16, 2009 12:39 AM
    Moderator