def main():
"""
Main routine: Cross reference a list of nodes to find common jobs
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[1:]
swfracs = list()
for date in joblist:
datelist = list()
datelist.append(date)
jobs = jobstats.alljobs(datelist)
corehrslt = 0.0
corehrsgt = 0.0
for job in jobs:
if 'sw' in job.queue:
if job.cores < 12:
corehrslt += job.cores*job.walltime
else:
corehrsgt += job.cores*job.walltime
if corehrslt > 0 and corehrsgt > 0:
frac = corehrslt/(corehrslt + corehrsgt)
print(date + " " + str(frac))
swfracs.append(frac)
print("===Mean===")
print(str(np.mean(swfracs)) + "+/-" + str(np.std(swfracs)/len(swfracs)))
return
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
jobs = jobstats.alljobs(joblist)
nodes = {}
for job in jobs:
if 'debug' not in job.queue and \
'lmgpu' not in job.queue and \
'scalemp' not in job.queue and \
'class' not in job.queue:
for node in job.nodes:
if node not in nodes:
nodes[node] = jobstats.nodeClass(job, node)
else:
nodes[node].joblist.append(job)
for node in nodes:
if len(nodes[node].joblist) > 20 and \
(nodes[node].avgEfficiency() < 0.1 or \
nodes[node].fractionBad() > 0.9):
nodes[node].printStats()
def main():
"""
Main routine: Print stats about ppn usage
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[1:]
usernames = {}
jobs = jobstats.alljobs(joblist)
for job in jobs:
if job.user in usernames:
usernames[job.user].addJob(job)
else:
usernames[job.user] = jobstats.userClass(job)
for user in sorted(usernames):
if any(1 < job.ppn < 12 for job in usernames[user].joblist):
print(user + ':')
usernames[user].printTopProp('ppn')
return
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 3:
sys.exit('Usage: ' + sys.argv[0] + ' username ' \
+ ' [Accounting files]')
else:
username = sys.argv[1]
joblist = sys.argv[2:]
user = None
jobs = jobstats.alljobs(joblist)
qtimes = list()
for job in jobs:
if username in job.user:
qtimes.append(job.tiq)
if user == None:
user = jobstats.userClass(job)
else:
user.addJob(job)
user.printAllJobs()
print("mean qtime: (" + str(np.mean(qtimes)/3600.0) + " +/- " \
+ str(np.std(qtimes)/(np.sqrt(len(qtimes))*3600.0)) + ") hours")
print("max qtime: " + str(max(qtimes)/3600.0) + " hours")
print("number of jobs: " + str(len(user.joblist)))
return
def main():
"""
Main routine: Show information about jobs with negative exit codes
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[1:]
probjobs = None
jobs = jobstats.alljobs(joblist)
for job in jobs:
if job.efficiency() > 10000.0:
print(str(job.id) + " " + str(job.efficiency()))
if job.exitcode < 0:
if probjobs != None:
probjobs.addJob(job)
else:
probjobs = jobstats.jobGroup(job)
probjobs.printStats()
print("Top Users:")
probjobs.printTopProp('user')
print("Top Nodes:")
probjobs.printTopProp('node')
print("Top Queues:")
probjobs.printTopProp('queue')
print("Top Exit Codes:")
probjobs.printTopProp('exitcode')
print("Super efficient Jobs:")
probjobs.printSuperEff()
jobs.printSuperEff()
return
def main():
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
jobs = jobstats.alljobs(joblist)
qts = list()
wallts = list()
corehrs = list()
allqts = list()
for job in jobs:
if job.cores > 1 and job.tiq > 0 and not np.isnan(job.tiq):
allqts.append(np.log10(job.tiq/3600.0))
corehrs.append(np.log10(job.cores*job.walltimereq/3600.0))
if job.cores == 1 and job.tiq > 0 and not np.isnan(job.tiq):
qts.append(np.log10(job.tiq/3600.0))
wallts.append(np.log10(job.walltimereq/3600.0))
print('max qt:' + str(max(qts)))
print('avg qt:' + str(np.mean(qts)) + '+/-' + str(np.std(qts)/np.sqrt(len(qts))))
subx = list()
suby = list()
for i,j in zip(wallts,qts):
if j > -1.0 and i > -1.0:
subx.append(i)
suby.append(j)
z = np.polyfit(subx,suby,1)
p = np.poly1d(z)
plt.cla()
plt.hexbin(wallts, qts, bins='log')
plt.plot([min(qts), max(qts)], [min(qts), max(qts)], 'k--')
plt.plot(subx, p(subx), 'k')
plt.xlabel('log(Walltime (hours))')
plt.ylabel('log(Queue Time (hours))')
plt.savefig('QueueTimes.png')
subx = list()
suby = list()
for i,j in zip(corehrs,allqts):
if j > -1.0 and i > -1.0:
subx.append(i)
suby.append(j)
z = np.polyfit(subx,suby,1)
p = np.poly1d(z)
plt.cla()
plt.hexbin(corehrs, allqts, bins='log')
plt.plot([min(allqts), max(allqts)], [min(allqts), max(allqts)], 'k--')
plt.plot(subx, p(subx),'k')
plt.xlabel('log(Core Hours)')
plt.ylabel('log(Queue Time (hours))')
plt.savefig('CHvsQT.png')
def main():
if len(sys.argv) < 2:
sys.exit("Usage: " + sys.argv[0] + " [Accounting files]")
else:
joblist = sys.argv[1:]
reset_time = dt.datetime(2013, 05, 10, 19, 30)
reset_time_unix = time.mktime(reset_time.timetuple())
crash_times = list()
crash_times.append(dt.datetime(2013, 05, 11, 0, 50))
crash_times.append(dt.datetime(2013, 05, 11, 7, 00))
crash_times.append(dt.datetime(2013, 05, 11, 7, 25))
crash_times.append(dt.datetime(2013, 05, 11, 8, 18))
crash_times.append(dt.datetime(2013, 05, 11, 8, 50))
crash_times.append(dt.datetime(2013, 05, 11, 10, 55))
crash_times.append(dt.datetime(2013, 05, 13, 2, 43))
crash_times.append(dt.datetime(2013, 05, 13, 6, 46))
crash_times.append(dt.datetime(2013, 05, 13, 7, 14))
crash_times.append(dt.datetime(2013, 05, 13, 11, 55))
crash_times_unix = list()
for ct in crash_times:
crash_times_unix.append(time.mktime(ct.timetuple()))
jobs = jobstats.alljobs(joblist)
ctimes = list()
etimes = list()
qtimes = list()
starts = list()
ends = list()
for job in jobs:
if job.ctime > reset_time_unix and job.start > 0:
ctimes.append(job.ctime)
etimes.append(job.etime)
qtimes.append(job.qtime)
starts.append(job.start)
ends.append(job.end)
plt.cla()
f, (ax1, ax2, ax3, ax4, ax5) = plt.subplots(5, sharex=True, sharey=False, figsize=(30, 15))
ax1.hist(ctimes, bins=(max(ctimes) - min(ctimes)) / 900, color="k", label="Created")
ax1.vlines(crash_times_unix, 0, ax1.get_ylim()[1], "b")
ax1.legend(loc=2, borderaxespad=0.0)
ax2.hist(etimes, bins=(max(etimes) - min(etimes)) / 900, color="k", label="Eligible")
ax2.vlines(crash_times_unix, 0, ax2.get_ylim()[1], "b")
ax2.legend(loc=2, borderaxespad=0.0)
ax3.hist(qtimes, bins=(max(qtimes) - min(qtimes)) / 900, color="k", label="Queued")
ax3.vlines(crash_times_unix, 0, ax3.get_ylim()[1], "b")
ax3.legend(loc=2, borderaxespad=0.0)
ax4.hist(starts, bins=(max(starts) - min(starts)) / 900, color="k", label="Started")
ax4.vlines(crash_times_unix, 0, ax4.get_ylim()[1], "b")
ax4.legend(loc=2, borderaxespad=0.0)
ax5.hist(ends, bins=(max(ends) - min(ends)) / 900, color="k", label="Finished")
ax5.vlines(crash_times_unix, 0, ax5.get_ylim()[1], "b")
ax5.legend(loc=2, borderaxespad=0.0)
plt.xlabel("Unix Epoch Time")
f.subplots_adjust(hspace=0)
plt.savefig("ctimes.png")
def main():
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
jobs = jobstats.alljobs(joblist)
qts = [job.tiq for job in jobs]
if len(qts) > 0:
meanqt = np.mean(qts)
if(meanqt) > 0.0:
errqt = np.std(qts)/np.sqrt(len(qts))
print(str(joblist).strip("[]'/tmp") + ' ' + str(meanqt) + ' ' + str(errqt))
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 3:
sys.exit("Usage: " + sys.argv[0] + " username " + " [Accounting files]")
else:
queue = sys.argv[1]
joblist = sys.argv[2:]
jobs = jobstats.alljobs(joblist)
qnum = 0
print("searching queue " + queue)
for job in jobs:
if queue in job.queue:
qnum += 1
print("Queue " + queue + " used in " + str(qnum) + " of " + str(len(jobs)) + " jobs")
return
def main():
"""
Main routine: Scans through jobs and identifies jobs that were running at the specified time
"""
#Parse command-line options
usage = "usage: %prog [options] [time (HH:MM:DD or Epoch time)] [Accounting files]"
parser = OptionParser(usage=usage)
parser.add_option("-n", "--node", dest="node", action="append",\
type="string", help="Search only for specified node(s)")
parser.add_option("-d", "--date", dest="date", action="store",\
type="string", help="Specify a date in YYYY-MM-DD format (default: today)")
parser.add_option("-s", "--summary", dest="summarize", action="store_true",\
default=False, help="Print a summary of each job (default: jobID only)")
(options, args) = parser.parse_args()
if len(args) < 2:
sys.exit(usage + '\n --help for list of options')
else:
time = gettime(args[0], options.date)
joblist = args[1:]
#Filters prevent storing undesired jobs in the jobs list
filters = None
if options.node != None:
filters = options.node
jobs = jobstats.alljobs(joblist, filters)
for job in jobs:
if job.wasRunningAt(time):
#Skip jobs on nodes that weren't specified
if options.node != None:
if not any(n in job.nodes for n in options.node):
continue
if options.summarize:
job.printStats()
else:
print job.id
return
def main():
"""
Main routine: Cross reference a list of nodes to find common jobs
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[2:]
jobs = jobstats.alljobs(joblist)
nodelist = ['sw-2r13-n50', 'sw-2r13-n51', 'sw-2r14-n20', 'sw-2r15-n17', 'sw-2r15-n44', 'sw-2r15-n66']
print('cross referencing nodes:')
print(nodelist)
for job in jobs:
if all(node in job.nodes for node in nodelist):
job.printStats()
return
def main():
"""
Main routine: Process total core days used by all jobs
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[1:]
jobs = jobstats.alljobs(joblist)
coredays = 0.0
for job in jobs:
coredays += job.cores*job.walltime
coredays /= 24.0*3600.0
print( '%.1f' %coredays)
return
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
corehours = np.zeros(1000)
jobs = jobstats.alljobs(joblist)
for job in jobs:
if job.cores < 1000:
corehours[job.cores] += job.cores*job.walltime/3600.0
plt.cla()
plt.bar(range(1,1001), list(corehours))
plt.xlabel('Cores')
plt.ylabel('Core hours')
plt.savefig('corehoursvprocs.png')
return
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 3:
sys.exit('Usage: ' + sys.argv[0] + ' username ' \
+ ' [Accounting files]')
else:
username = sys.argv[1]
joblist = sys.argv[2:]
user = None
jobs = jobstats.alljobs(joblist)
for job in jobs:
if username in job.user:
if user == None:
user = jobstats.user(job)
else:
user.addJob(job)
user.printStats()
return
def main():
"""
Main routine: Print User stats
"""
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] \
+ ' [Accounting files]')
else:
joblist = sys.argv[1:]
exitCodes = {}
jobs = jobstats.alljobs(joblist)
for job in jobs:
if job.exitcode not in exitCodes:
exitCodes[job.exitcode] = 1
else:
exitCodes[job.exitcode] += 1
sortedecs = sorted(exitCodes.iteritems(), key=operator.itemgetter(1))
for ec, quant in sortedecs:
print ec, 100.0*quant/len(jobs)
return
def main():
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
jobs = jobstats.alljobs(joblist)
qts_ppn = list()
qts_procs = list()
corehrs_ppn = list()
corehrs_procs = list()
for job in jobs:
if job.cores > 1 and job.tiq > 0 and not np.isnan(job.tiq):
if job.ppn == 0:
qts_procs.append(np.log10(job.tiq/3600.0))
corehrs_procs.append(np.log10(job.cores*job.walltimereq/3600.0))
else:
qts_ppn.append(np.log10(job.tiq/3600.0))
corehrs_ppn.append(np.log10(job.cores*job.walltimereq/3600.0))
subx_procs = list()
suby = list()
for i,j in zip(corehrs_procs,qts_procs):
if j > -1.0 and i > -1.0:
subx_procs.append(i)
suby.append(j)
z_procs = np.polyfit(subx_procs,suby,1)
p_procs = np.poly1d(z_procs)
print(z_procs)
subx_ppn = list()
suby = list()
for i,j in zip(corehrs_ppn, qts_ppn):
if j > -1.0 and i > -1.0:
subx_ppn.append(i)
suby.append(j)
z_ppn = np.polyfit(subx_ppn,suby,1)
p_ppn = np.poly1d(z_ppn)
print(z_ppn)
plt.cla()
plt.hexbin(corehrs_procs, qts_procs, bins='log')
#plt.plot([min(corehrs_procs), max(corehrs_procs)], [min(qts_procs), max(qts_procs)], 'k--')
plt.plot(subx_ppn, p_ppn(subx_ppn), 'y', linewidth=4)
plt.plot(subx_procs, p_procs(subx_procs), 'w', linewidth=4)
plt.xlabel('log(Core hours requested)')
plt.ylabel('log(Queue Time (hours))')
plt.savefig('procs.png')
plt.cla()
plt.hexbin(corehrs_ppn, qts_ppn, bins='log')
# plt.plot([min(qts_ppn), max(qts_ppn)], [min(qts_ppn), max(qts_ppn)], 'k--')
plt.plot(subx_procs, p_procs(subx_procs), 'w', linewidth=4)
plt.plot(subx_ppn, p_ppn(subx_ppn),'y', linewidth=4)
plt.xlabel('log(Core hours requested)')
plt.ylabel('log(Queue Time (hours))')
plt.savefig('ppn.png')
def main():
if len(sys.argv) < 2:
sys.exit('Usage: ' + sys.argv[0] + ' [Accounting files]')
else:
joblist = sys.argv[1:]
reset_time = dt.datetime(2013, 05, 10, 19, 30)
reset_time_unix = time.mktime(reset_time.timetuple())
crash_times = list()
crash_times.append(dt.datetime(2013, 05, 11, 0, 50))
crash_times.append(dt.datetime(2013, 05, 11, 7, 00))
crash_times.append(dt.datetime(2013, 05, 11, 7, 25))
crash_times.append(dt.datetime(2013, 05, 11, 8, 18))
crash_times.append(dt.datetime(2013, 05, 11, 8, 50))
crash_times.append(dt.datetime(2013, 05, 11, 10, 55))
crash_times.append(dt.datetime(2013, 05, 13, 2, 43))
crash_times.append(dt.datetime(2013, 05, 13, 6, 46))
crash_times.append(dt.datetime(2013, 05, 13, 7, 14))
crash_times.append(dt.datetime(2013, 05, 13, 11, 31))
crash_times.append(dt.datetime(2013, 05, 13, 11, 41))
crash_times.append(dt.datetime(2013, 05, 13, 11, 55))
crash_times_unix = list()
for ct in crash_times:
crash_times_unix.append(time.mktime(ct.timetuple()))
jobs = jobstats.alljobs(joblist)
ctimes = list()
etimes = list()
qtimes = list()
starts = list()
ends = list()
for job in jobs:
if job.ctime > reset_time_unix and job.start > 0:
print(job.start)
if job.ctime < 0 or job.etime < 0 or job.qtime < 0 or job.start < 0 or job.end < 0:
print(job.id)
ctimes.append(job.ctime)
etimes.append(job.etime)
qtimes.append(job.qtime)
starts.append(job.start)
ends.append(job.end)
plt.cla()
f, (ax1, ax2, ax3, ax4, ax5) = plt.subplots(5, sharex=True, sharey=False, figsize=(20, 10))
ax1.hist(ctimes, bins=(max(ctimes) - min(ctimes))/900, color='k', label='Created')
ax1.vlines(crash_times_unix, 0, ax1.get_ylim()[1], 'b')
ax1.legend(bbox_to_anchor=(1, 1), loc=2, borderaxespad=0.)
ax2.hist(etimes, bins=(max(etimes) - min(etimes))/900, color='k', label='Eligible')
ax2.vlines(crash_times_unix, 0, ax2.get_ylim()[1], 'b')
ax2.legend(bbox_to_anchor=(1, 1), loc=2, borderaxespad=0.)
ax3.hist(qtimes, bins=(max(qtimes) - min(qtimes))/900, color='k', label='Queued')
ax3.vlines(crash_times_unix, 0, ax3.get_ylim()[1], 'b')
ax3.legend(bbox_to_anchor=(1, 1), loc=2, borderaxespad=0.)
ax4.hist(starts, bins=(max(starts) - min(starts))/900, color='k', label='Started')
ax4.vlines(crash_times_unix, 0, ax4.get_ylim()[1], 'b')
ax4.legend(bbox_to_anchor=(1, 1), loc=2, borderaxespad=0.)
ax5.hist(ends, bins=(max(ends) - min(ends))/900, color='k', label='Finished')
ax5.vlines(crash_times_unix, 0, ax5.get_ylim()[1], 'b')
ax5.legend(bbox_to_anchor=(1, 1), loc=2, borderaxespad=0.)
plt.xlabel('Unix Epoch Time')
f.subplots_adjust(hspace=0, right=0.8)
plt.savefig('ctimes.temp.png')
