Notes - Astro250/Sping2017 GitHub Wiki

Here I will post useful pointers for you at NERSC.

SSH

OK, for those of you who are not so familiar with ssh, here is some good material to have a look at. I used keychain to only have to load up your keys once at login to a shell window and then you never type it again. This saves you the trouble of doing the ssh-agent and ssh-add yourself. They refer to in this write-up above. The latest version of Mac OS X came out with their own thing in which they autoload your ssh keys for you.

HPCPorts

For those of you into either CMB and/or astro image processing, I highly recommend Ted Kisner's HPCPorts modules at NERSC. You can read up about it here.

NX

Also, you folks should be using NX to accelerate your X-Windows to NERSC and to keep a virtual desktop open while you work on things over days. Go to the NERSC homepage and log in in the upper right with your nim account and passwd. Then you can go to this page and follow the instructions for downloading, installing and using NX.

BASH & NIM

An important part of today's class on BASH programming is that everyone is using the bash shell. Please make sure to log into NIM, then go to ACTIONS on the top right, and select Change Shell. In the next window change everything to bash and save it.

Also, as we are going to start this week doing things on the NERSC machines, please realize that they often go up and down both predictably and unpredictably. You can find out the status on this page.

MPI Codes

All the MPI codes have been compiled successfully in /project/projectdirs/astro250/class/code/mpi Copy them over and you can start playing with them.

Job Script on NERSC

Please have a look at the NERSC web-pages on how to run in different configurations for the hybrid codes. They can be found here.

Also, for now I would use the job script generator at NERSC. I have a ticket in with them as to why it behaves the way it does, but for now I would trust them that they set it up right, even though the "c" value is always maxed out. This is a rather interesting question they are investigating as we speak.

Remember that if you start playing with the KNL's you might wish to have no more than 64 cores running. You can reserve 2 for the OS using #SBATCH -S 2.

Class Reservation

Our reservations have been created -- they are called them astro250m and astro250w (for Monday and Wednesday respectively). To use the reservation you just have to add --reservation=astro250m either on the job submit line or in the script as one of the #SBATCH directives.

salloc -N 1 -p debug -t 00:30:00 -C haswell --account=m2218 --reservation=astro250m

Energy Used on NERSC

And here is the cool command to get your energy usage on cori. myjobid, is just the SLURM jobid of your run.

sacct -j myjobid --noconvert -o jobid,JobName,NNodes,ConsumedEnergyRAW,Elapsed