[premise-users] Premise nodes available and pending jobs

[Tavares De Oliveira Melo, Arthur]

Hi,

Thank you MacManes for the input and thank you Tony for the explanation. You're right.

I have requested a GPU, simply because I copy the sbatch command from the SLURM link. Sorry about this. Indeed, I'm not sure what this GPU means and there are some advantage in request it? I wondering to remove this request from my command (attached).

Thank you again.

Best,
Arthur



Arthur Melo
Bioinformatician Postdoctoral Researcher
University of New Hampshire | College of Life Sciences and Agriculture
http://www.unh.edu/halelab
________________________________
From: Anthony Westbrook <anthonyw at wildcats.unh.edu>
Sent: Saturday, June 3, 2017 1:06 PM
To: Tavares De Oliveira Melo, Arthur; premise-users at lists.sr.unh.edu
Subject: RE: Premise nodes available and pending jobs

Hi Arthur -

Queuing can always be a little tricky - but this is good info to recap for everyone.  While the group who was responsible for purchase will have higher priority on their nodes, all nodes are still available to everyone.  If there are free nodes and no jobs waiting for them, anyone's job will be dispatched to them, regardless of the user's group.

If your job is being queued, it means there is currently a resource not available (like Matt mentioned).  It can be something like:


  1.  You requested more nodes than are currently available (e.g. your job requires 4 nodes, but only 2 are available)
  2.  You requested more memory than any free node currently has (12 nodes have 128GB, 4 have 512GB. So if you request 400GB of RAM and all 4 high-memory nodes are unavailable, your job will queue)
  3.  You requested a GPU.  Similar to the high memory nodes, only 4 nodes have GPUs in them.

In your particular case, I took a look at one of your Slurm scripts, and it looks like you were requesting a GPU (which are currently in use).

Let me know if you have further questions too.  Thanks -

Toni Westbrook
Computational Scientist
Research Computing Center, College of Life Sciences and Agriculture
University of New Hampshire
Office: 436 Gregg Hall



From: premise-users [mailto:premise-users-bounces at lists.sr.unh.edu] On Behalf Of Tavares De Oliveira Melo, Arthur
Sent: Friday, June 02, 2017 1:37 PM
To: premise-users at lists.sr.unh.edu
Subject: [premise-users] Premise nodes available and pending jobs

Hi Tony and all Premise users,

I hope this message finds you well.

I was wondering how the premise nodes distribution works. I mean, out of 14 nodes available at premise, there are an amount specific for COLSA users? I'm asking because today morning I tried to launch a job (#6914) and until now the status is PD (pending) and the reason is "Resources". If I'm correct it means the job is waiting for computer resource. Rigth? Also, using "squeue" to monitoring the job I can see only 8 nodes being use. Initialy I supposed the COLSA users can not access the entire premise resources available (14 nodes). What is not clear to me is yesterday I submitted a job (#6867) and it runned and completed normaly. So, I'm not unerstanding why my job #6914 is pending since 9 am.

Thank you very much.


Best,

Arthur


Arthur Melo
Bioinformatician Postdoctoral Researcher
University of New Hampshire | College of Life Sciences and Agriculture
http://www.unh.edu/halelab
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sr.unh.edu/pipermail/premise-users/attachments/20170605/a40d7407/attachment-0001.html>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: Arthur_sbacth.txt
URL: <http://lists.sr.unh.edu/pipermail/premise-users/attachments/20170605/a40d7407/Arthur_sbacth-0001.txt>