Hi Dave, I can confirm that CoreSpecCount can not be reset to 0 once it is set >0 (at least for FastSchedule>0). As a workaround for this bug you can try to stop slurmctld, remove node_state file and start slurmctld again.
Best regards, Taras On Fri, Aug 9, 2019 at 11:54 PM Guertin, David S. <[email protected]> wrote: > What's even stranger is that I can change CoreSpecCount to any other > number (2, 3, whatever), restart the daemons, and the change is made. But > if I try to set it to 0, nothing happens. It's like there's a setting > somewhere that is forcing CoreSpecCount to be >0. But I have no idea what > that could be. > > Dave > > *David Guertin* > *Information Technology Services* > *Middlebury College* > *700 Exchange St.* > *Middlebury, VT 05753* > *(802)443-3143* > ------------------------------ > *From:* slurm-users <[email protected]> on behalf of > Guertin, David S. <[email protected]> > *Sent:* Friday, August 9, 2019 4:28 PM > *To:* Slurm User Community List <[email protected]> > *Subject:* Re: [slurm-users] How to turn off core specialization? > > > Have you restarted all your slurm daemons? > > Yes, I have done that on every node, but it still shows one specialized > core. > > > Not sure whether "scontrol reconfigure" can deal with that change. > > I tried "scontrol reconfigure", but it also had no effect. > > Thanks, > Dave > >
