Re: using SQL for multi-machine job management?

On Sat, Sep 12, 2009 at 5:05 PM, jebjeb <martin(dot)belleau(at)yahoo(dot)com> wrote:

>
> I'm considering using PostgreSQL as part of the implementation of a
> multi-machine job management system. Here is an overview of the system:
>
> -jobs are submitted by an API and stored to a SQL database. Jobs contain a
> list of source filenames and a description of the operations to perform on
> the files (compress file, add it to an archive, encrypt it, compare with
> other file, etc).
>
> -multiple machines (up to 50?) look at the database and grabs a job. It
> will update the database to indicate that it will be the machine running
> this job. It will also update the database with the current completion
> progress (%) of this job.
>
> 1)Is this something done often using SQL databases?
> 2)The jobs will be quite CPU intensive: will I run into trouble if the
> database is located on one of the machine which will be executing the jobs?
> 3)I would like to have a backup ready to take over if the machine with the
> database fails. Only some of the info I store (data about a job, machine
> that is executing the job) is important to back up. Things like job
> progress doesn't have to be backed up. Any tips on how I should set up the
> database to accomplish this?
>
>
You might want to look into using something like SLURM or SGE or Condor for
doing this type of thing.

Sean