[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [HTCondor-users] Dynamic slots and concurrency...

"HTCondor-users" <htcondor-users-bounces@xxxxxxxxxxx> wrote on 11/05/2015 06:27:03 PM:

> From: Todd Tannenbaum <tannenba@xxxxxxxxxxx>

> To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
> Date: 11/05/2015 06:28 PM
> Subject: Re: [HTCondor-users] Dynamic slots and concurrency...
> Sent by: "HTCondor-users" <htcondor-users-bounces@xxxxxxxxxxx>
> On 11/5/2015 3:40 PM, Jody Pearson wrote:
> > Hi all,
> >
> > I'm wondering if concurrency limits are supposed to word with dynamic
> > slots ?
> >
> Yes, BUT you will likely want to configure HTCondor to use "Consumption
> Policies", which are not enabled by default.  See
> for more information.

This puzzles me, since as far as I can tell over the past three years, my dynamic slots with partitionable leftover claiming have not run into any oversubscription problems with the MATLAB license concurrency limit. I did try switching to consumption policies when they first came out, but ran into what must have been bug 4945 where claimed slots hung in the Matched state, so I went back to CPL.

I suppose I should take a closer look at this. What's the right way to determine the negotiator's current tally for a given concurrency limit? Could my lack of noticeable problems be a result of using a single centralized scheduler  via SCHEDD_HOST = <hostname> instead of per-machine schedulers?

-Michael Pelletier.