================================================================
domino.ucr.edu README
================================================================

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
muffin DOCUMENTATION
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Documentation for muffin can be found here:

https://github.com/derpycode/muffindoc

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
USER ACCOUNT ENVIRONMENT CONFIGURATION
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

>>> Add the following text in:

.bash_profile

(located in your home directory) and under (and replacing all text following it) the comment line:

# User specific environment and startup programs

PATH=$PATH:$HOME/bin
export PATH
PATH=$PATH:/state/partition1/apps/netcdf/lib
export PATH
PATH=$PATH:/state/partition1/apps/netcdf/include
export PATH
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/state/partition1/apps/netcdf/lib
export LD_LIBRARY_PATH
export CGENIE_PYTHON=/state/partition1/apps/Python-2.7.13/bin/python2.7

>>> Once you have done this and saved the file, log out and re-log in again.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
muffin INSTALLATION AND CONFIGURATION
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

>>> install GENIE:

$ git clone https://github.com/derpycode/cgenie.muffin.git

>>> The only configuration edit you need make is to uncomment the 'domino' setting in:

user.mak 

>> and comment out the default line, which can all be found at the bottom of the file below the comment/header line:

# === NetCDF library ===

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
CLUSTER QUEUE USAGE
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

There are 2 seperate queues on domino.ucr.edu:

dog.q ----- 4 nodes of 8 cores each == 32 cores total
puppy.q --- 4 (soon to be 8) nodes of 12 cores each == 48 (soon 96) cores total

USAGE:

dog.q ----- for generic, random, test and relatively short (< 1 month) runs 
puppy.q --- for experiment ensembles

please sick to this usage as far as possible.
The aim is to enable large ensembles to be run efficiently as and when they need to be run try
by keeping puppy.q relatively empty if possible (and dog.q full)
If dog.q is full, then submit to puppy.q

For small numbers (ideally <=8) of very long runs (>>1 month), there
is a dedicated queue (kitten.q) on eevee.ucr.edu (see me for a seperate account).

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
JOB QUEUE MANAGEMENT
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

>>> queue listing for own user jobs:
$ qstat -f
>>> queue listing for all user jobs:
qstat -f -u "*"

>>> submit job to dog.q queue:
$ qsub -q dog.q -j y -o cgenie_log -V -S /bin/bash
>>> submit job to puppy.q queue:
$ qsub -q puppy.q -j y -o cgenie_log -V -S /bin/bash

>>> delete job number XXX
$ qdel XXX
>>> deletes all your jobs
$ qdel *