[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Hooks in the Scheduler Universe?
- Date: Wed, 6 Oct 2010 17:37:46 -0400
- From: Ian Chesal <ichesal@xxxxxxxxxxxxxxxxxx>
- Subject: Re: [Condor-users] Hooks in the Scheduler Universe?
On Wed, Oct 6, 2010 at 3:03 PM, Matthew Farrellee <matt@xxxxxxxxxx> wrote:
Ian Chesal wrote:
That pipe/signal bug never got fully flushed out did it?
On Wed, Oct 6, 2010 at 10:28 AM, David Arthur <mumrah@xxxxxxxxx <mailto:mumrah@xxxxxxxxx>> wrote:
I'm hoping to use the Job Hook system for DAGs. I would like to
utilize the "prepare job" and "job exit" hooks (to get notifications
when the DAG is about to start, and once it's complete).
I don't want to knock too much on hooks, but unless you're using them on non-Windows nodes I wouldn't recommend this approach. The technology is still unstable on Windows. It's very easy to crash a startd.
I've used this for VM image transfer and cache maintenance. It is quite useful to realize that hooks can run for any job independently of if the job was pulled from the FETCH_WORK hook.
I've never considered running *just* the prepare_job hook, without a corresponding fetch_work hook.
That's a nice bit of outside thinking that is. Now to get it working on Windows... :)