Re: [HTCondor-users] Minimize time to start job.

[John M Knoeller <johnkn@xxxxxxxxxxx>]

Setting the cycle delay to 0 puts more of a load on the Collector, since the Negotiator will be fetching all of the Machine ads and Submitter ads more often, but if your Collector is keeping up, then this is fine.

 

Keep in mind that having the negotiation cycle be continuous is wasted effort once all of your machines are busy running jobs.

 

The normal state of an HTCondor pool is nearly all of the execute nodes busy, and an infinite supply of jobs in the Schedd waiting for their turn,

this is what we tune for â not the startup time of an empty pool.   Hence the 20 second cooldown between negotiation cycles.

 

-tj

 

From: HTCondor-users <htcondor-users-bounces@xxxxxxxxxxx> On Behalf Of don_vanchos
Sent: Tuesday, August 13, 2019 6:21 AM
To: HTCondor-Users Mail List <htcondor-users@xxxxxxxxxxx>
Subject: [HTCondor-users] Minimize time to start job.

 

Hello,

 

I noticed that for the simplest vanilla jobs on my cluster, the time difference `job_ad["JobStartDate"] - job_ad["QDate"]` is from 5 to 20 seconds. So quite a lot of time elapses between sending job to the queue and starting the process.

 

Then I set NEGOTIATOR_CYCLE_DELAY setting to 0. And this time difference became equal from 0 to 1 second.

 

My goal is to make job launch as fast as possible! What are the consequences if I make this setting equal to zero? Maybe performance degradation? Or maybe the wrong behavior in some cases? If the '0' value is harmful, then how can I minimize this time difference?

 

P.S. Many thanks to all for the answers to my questions in the neighboring branches and in this (in advance).

 

--

Sincerely yours,
Ivan Ergunov                                                 mailto:hozblok@xxxxxxxxx