Skip to content

Instantly share code, notes, and snippets.

@oweidner
Created February 7, 2012 00:40
Show Gist options
  • Save oweidner/1756215 to your computer and use it in GitHub Desktop.
Save oweidner/1756215 to your computer and use it in GitHub Desktop.
BigJob/Bliss Echo Test (PhD-Research/1styearrev/experiments/01_bfast_benchmark)
""" Example application demonstrating how to submit a simple
/bin/echo job with BigJob.
"""
import os
import time
import pdb
import sys
sys.path.insert(0, os.getcwd() + "/../")
COORDINATION_URL = "redis://[email protected]:6379"
RESOURCEMGR_URL = "pbs+ssh://localhost"
from bigjob import bigjob, subjob, description
### This is the number of jobs you want to run
NUMBER_JOBS=1
def has_finished(state):
state = state.lower()
if state=="done" or state=="failed" or state=="canceled":
return True
else:
return False
if __name__ == "__main__":
starttime=time.time()
##########################################################################################
# make sure you are familiar with the queue structure on lonestar, your project id
# and the walltime limits on each queue. change accordingly
#
queue=None # You can switch to development for very short runs
project=None # since FutureGrid is an experimental infrastrucutre.
walltime=60 # Time in minutes. There are limits on the time you can request.
processes_per_node=8 # if you want more memory per process you can reduce this to 6, 2 or 1
number_of_processes=24 # The total number of processes you want running, depends on the value above
# yye00: to keep things clean: create the directory agent where you intend to run the script
# you can start by doing this in SCRATCH if you run out of space. This directory
# will get big quickly so keep an eye out and do not put it in $HOME
workingdirectory= os.path.join("/N/u/oweidner/tmp") # working directory for agent.
##########################################################################################
print "Start Pilot Job/BigJob at: " + RESOURCEMGR_URL
bj = bigjob(COORDINATION_URL)
bj.start_pilot_job( RESOURCEMGR_URL,
None,
number_of_processes,
queue,
project,
workingdirectory,
None,
walltime,
processes_per_node)
print "Pilot Job/BigJob URL: " + bj.pilot_url + " State: " + str(bj.get_state())
##########################################################################################
# Submit jobs through BigJob
# Here you can add any arguments to each SubJob, change the ouput and error filenames and so on
# change this to your heart's content, but be careful
jobs = []
job_start_times = {}
job_states = {}
for i in range(0, NUMBER_JOBS):
jd = description()
jd.executable = "/bin/echo"
jd.number_of_processes = "1"
jd.spmd_variation = "mpi"
jd.arguments = ["Hello from BigJob"]
jd.output = "sj-stdout-"+str(i)+".txt"
jd.error = "sj-stderr-"+str(i)+".txt"
sj = subjob()
sj.submit_job(bj.pilot_url, jd)
jobs.append(sj)
job_start_times[sj]=time.time()
job_states[sj] = sj.get_state()
# busy wait for completion
while 1:
finish_counter=0
result_map = {}
for i in range(0, NUMBER_JOBS):
old_state = job_states[jobs[i]]
state = jobs[i].get_state()
#print "Job " + str(jobs[i]) + " state: " + state
if result_map.has_key(state)==False:
result_map[state]=1
else:
result_map[state] = result_map[state]+1
#pdb.set_trace()
if old_state != state:
print "Job " + str(jobs[i]) + " changed from: " + old_state + " to " + state
if old_state != state and has_finished(state)==True:
print "Job: " + str(jobs[i]) + " Runtime: " + str(time.time()-job_start_times[jobs[i]]) + " s."
if has_finished(state)==True:
finish_counter = finish_counter + 1
job_states[jobs[i]]=state
if finish_counter == NUMBER_JOBS:
break
time.sleep(2)
runtime = time.time()-starttime
print "Runtime: " + str(runtime) + " s; Runtime per Job: " + str(runtime/NUMBER_JOBS)
##########################################################################################
# Cleanup - stop BigJob
bj.cancel()
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment