Hi,
Some of my jobs stay in IDLE for a really
long time compared to what I have configured.
On the submit node, I set
NEGOTIATOR_INTERVAL to 5 seconds, expecting jobs to start
extremely quickly after being submitted.
This is a local pool for test and training
purposes, so I actually want the jobs to start and execute
almost immediately.
When looking at the NegotiatorLog file, it
looks like the negotiation cycles are happening as often as
intended, yet most of my jobs stay in queue for 5 minutes.
Those are docker universe jobs. Condor_q
-better-analyze tells me there are machines capable of running
my jobs.
Even weirder, I tried to download one of my
docker images as a TAR archive and load manually in Docker,
and then the jobs using that image start immediately.
The same jobs using the same image but from
a Docker repo take 5 minutes to start.
Is there another knob I could turn to speed
up the process ?
![[Computer Systems Lab]](https://www-auth.cs.wisc.edu/pics/csl_logo.gif){kind=logo}