Ok I'll try this tomorrow.

Thanks
Jackie

Sent from my iPhone

> On May 26, 2014, at 5:54 AM, Dominik Bartkiewicz <[email protected]> wrote:
>
>
>> On 05/24/2014 01:15 AM, Jacqueline Scoggins wrote:
>> I will definitely make sure the nodes have the same slurm.conf file. We
>> use warewulf to update those and then I do scontrol reconfig and that
>> works.  My concerns is the tmp files for running and pending jobs when I
>> stop and restart the service that they come back up without any problems
>> and definitely don't get lost.
>>
>> Thanks
>>
>> Jackie
>>
>>
>> On Fri, May 23, 2014 at 2:35 PM, David Bigagli <[email protected]
>> <mailto:[email protected]>> wrote:
>>
>>
>>
>>    Make sure you copy the subdirectories job.n with the job files and
>>    job environment. If you have a backup slurmctld stop it as well.
>>
>>    Since you change the slurm.conf at the end you may want to restart
>>    the slurmd as well so they will have the same configuration as the
>>    controller otherwise you will get the warning in the controller log
>>    about slurmd running with different configuration.
>>
>>
>>    On 05/23/2014 02:08 PM, Jacqueline Scoggins wrote:
>>
>>        Is there a clean way for me to change the SlurmdSpoolDir in
>>        slurm.conf
>>        without impacting the running jobs.  I would like move it to a
>>        different
>>        location but would like for all jobs in the queue to be fine. I will
>>        stop slurm,  move the files in the existing location and restart
>>        slurm.
>>           Will this be a problem with the jobs that are currently
>>        running or queued?
>>
>>        If there is a better way of doing this please let me know.
>>
>>        Thanks
>>
>>        Jackie
>>
>>
>>    --
>>
>>    Thanks,
>>           /David/Bigagli
>>
>>    www.schedmd.com <http://www.schedmd.com>
>
>
> We have used this procedure:
>
> change slurm.conf
> restart slurmctl
> stop slurmd
> rsync old SlurmSpoolDir to new localization
> replaced all socket in new localization by symbolic link to old socket
> start slurmd
>
>
> DB

Reply via email to