nl.nikhef.ndpf.groupviews/trunk/ndpf-gv-mkplots

#! /usr/bin/env python
# $Id$
# Source: $URL$
# J. A. Templon, NIKHEF/PDP 2011

import optparse

p = optparse.OptionParser(description="Program to make rrdtool plots " + \
                          "of job running jobs by unix group")

# p.add_option("-r",action="store",dest="minsize",default='0',help="minimum size of dirs considered; can use suffixes k,M,G for multiples of 1000**{1,2,3} bytes")
# p.add_option("--qdel",action="store_true",dest="deljobs",help="delete jobs for which TMPDIR is larger than MINSIZE",default=False)

p.add_option("--rank-only",action="store_true",dest="rankonly",
             help="don't plot, just print ranking of groups",default=False)

debug = 0

opts, args = p.parse_args()

import os

DATADIR=os.environ['HOME'] + '/ndpfdata/'
PLOTDIR='/www/grid/stats/groupviews/ndpf/'

# 8 class qualitative paired color scheme

colors = [ "#A6CEE3",  "#1F77B4",  "#B2DF8A",  "#33A02C",
           "#FB9A99",  "#E31A1C",  "#FDBF6F",  "#FF7F00"    ]

colors.reverse()

# for reference : ranges of the RRAs
# base step size 60 sec
# step  1 :  60 sec x 1600 points : 1600 min : 26,67 hr : 1.11 days
# step  2 : 120 sec x 1200 points : 2400 min : 40    hr : 1,67 days
# step 10 : 600 sec x 1800 points : 18 000 min : 300 hr : 12,5 days
# step 30 : 1800 sec x 2500 points : 75 000 min : 1250 hr : 52.08 days
# step 120 : 7200 sec x 1000 points : 120 000 min : 2000 hr : 83.33 days
# step 480 : 28800 sec x 1000 points : 480 000 min : 8000 hr : 333.33 days
# step 1440 : 86400 sec x 3650 points : 3650 days : 10 years

# resolutions of RRAs

# 1 :     60 sec :    1 min
# 2 :    120 sec :    2 min
# 3 :    600 sec :   10 min
# 4 :  1 800 sec :   30 min
# 5 :  7 200 sec :  120 min :  2 hr
# 6 : 28 800 sec :  480 min :  8 hr
# 7 : 86 400 sec : 1440 min : 24 hr : 1 day

NUMQS = 33                      # number of quarters for "alltime" plot

plotrangedef = {
    'hr' : { 'timeargs' : [ '-s', 'n-200min',  '-e', 'n' ],
             'timetag'  : 'hr',
             'avrange' : 3*3600,
             'avres'   : 60,
             'sizeargs' : { 'small' : [ '--width', '200', '--height', '125',
                                        '--x-grid',
                                        'MINUTE:20:HOUR:1:HOUR:1:0:%H:00'
                                        ],
                            'large' : [ '--width', '800', '--height', '500' ]
                            },
             'numgroups' : 4,
             },
    'day' : { 'timeargs' : [ '-s', 'n-2000min',  '-e', 'n' ],
              'timetag' : 'day',
             'avrange' : 24*3600,
             'avres'   : 60,
             'sizeargs' : { 'small' : [ '--width',  '200', '--height', '125',
                                        '--x-grid',
                                        'HOUR:6:DAY:1:HOUR:12:0:%a %H:00'
                                        ],
                            'large' : [ '--width', '1000', '--height', '625' ]
                            },
              'numgroups' : 5,
              },
    'week' : { 'timeargs' : [ '-s', 'n-288hr',  '-e', 'n' ],
               'timetag' : 'week',
               'avrange' : 7*24*3600,
               'avres'   : 600,
               'sizeargs' : { 'small' : [ '--width', '576', '--height', '125' ],
                              'large' : [ '--width', '1728', '--height', '375',
                                          '-n', 'DEFAULT:16:']
                              },
               'numgroups' : 6,
               },
    
    # note the construction "repr(576*n)" here --- this is because the plot is
    # (a multiple of) 576 pixels, and 120 min is one of the RRAs, so choosing
    # a lower limit of 576*120 gives us a plot with one pixel per RRA bin.
    
    'month' : { 'timeargs' : [ '-s', 'n-'+repr(576*120)+'min',  '-e', 'n'],
               'timetag' : 'month',
               'avrange' : 31*24*3600,
               'avres'   : 1800,
               'sizeargs' : { 'small' : [ '--width', '576', '--height', '105'],
                              'large' : [ '--width', '2304', '--height', '420',
                                          '-n', 'DEFAULT:18:',
                                          '--x-grid',
                                          'HOUR:12:DAY:1:DAY:3:86400:%d-%b'
                                          ]
                              },
               'numgroups' : 7,
                },

    'year' : { 'timeargs' : [ '-s', 'n-'+repr(576*1440)+'min',  '-e', 'n'],
               'timetag' : 'year',
               'avrange' : 365*24*3600,
               'avres'   : 86400,
               'sizeargs' : { 'small' : [ '--width', '576', '--height', '105'],
                              'large' : [ '--width', '2304', '--height', '420']
                              },
               'numgroups' : 8,
                },

# adjusting for nice plot ... set NUMQS above; range is NUMQS/4 years
    
    'alltime' : { 'timeargs' : [ '-s', 'n-'+repr(NUMQS*365*1440/4)+'min',  '-e', 'n'],
                  'timetag' : 'alltime',
                  'avrange' : NUMQS*365*24*3600/4,
                  'avres'   : 86400,
                  'sizeargs' : { 'small' : [ '--width', '576', '--height', '105',
                                             '--x-grid',
                                             'MONTH:3:YEAR:1:YEAR:1:31536000:%Y'
                                             ],
                                'large' : [ '--width', '2304', '--height', '420',
                                            '--x-grid',
                                            'MONTH:1:YEAR:1:MONTH:3:2592000:%b-%Y'
                                            ]
                                },
               'numgroups' : 8,
                },

                  
}

commonargs = ['--imgformat', 'PNG', 
              '--legend-position=east', '--legend-direction=bottomup']

import rrdtool
import time
import glob

### function definitions

def doplot(glist, dbtype, psize, timetag, sizeargs, timeargs, pcents, ranktype):

    grouplist = glist[:]
    defs = list()
    plots = list()

    data_defs = list()
    plot_defs = list()

    gcolors = dict()
    
    skipgroup = 0
    if ranktype == 'bottom':
        if 'unused'  in grouplist:
            grouplist.remove('unused')
            grouplist.insert(0,'unused')
        if 'offline'  in grouplist:
            grouplist.remove('offline')
            grouplist.insert(0,'offline')
    else:
        if 'unused'  in grouplist:
            skipgroup += 1
        if 'offline'  in grouplist:
            skipgroup += 1

    for idx in range(len(grouplist)):
        thisgrp = grouplist[idx]
        if thisgrp == 'unused':
            gcolors[thisgrp] = '#d8d8d8'
        elif thisgrp == "offline":
            gcolors[thisgrp] = "#790ead"
        else:
            gcolors[thisgrp] = colors[idx+skipgroup]

    if dbtype == 'queued':
        if 'unused'  in grouplist: grouplist.remove('unused')
        if 'offline' in grouplist: grouplist.remove('offline')
    
    if ranktype == 'top':
        for group in (grouplist + ['total']):
            data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')
        otherstring = 'CDEF:other=total,'
        for group in grouplist:
            otherstring += group + ','
        otherstring += (len(grouplist)-1) * '+,' + '-'

#        print otherstring
        data_defs.append(otherstring)
    elif ranktype == 'bottom':
        for group in grouplist:
            data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')
    else:
        print 'Unknown ranktype detected:', ranktype
        sys.exit(2)
    
    sumshown = 0
    for idx in range(len(grouplist)):
        group = grouplist[idx]
        acolor = gcolors[group]
        pdefstr = 'AREA' ':' + group + acolor + ':' + "%8s" % (group)
        if pcents:
            pdefstr = pdefstr + ' (' + "%4.1f" % (pcents[group]) + ')'
            sumshown += float(pcents[group])
        pdefstr = pdefstr + '\\n'
        pdefstr = pdefstr + ':STACK'
        plot_defs.append(pdefstr)

    if ranktype == 'top':
        pdefstr = 'AREA' ':' + 'other' + '#794044' + ':' + '   other'
        if pcents:
            pdefstr = pdefstr + ' (' + "%4.1f" % (100 - sumshown) + ')'
        pdefstr = pdefstr + '\\n'
        plot_defs.insert(0,pdefstr)
        plot_defs.append("LINE:total#000000") # :total")

    pargs = [ PLOTDIR + dbtype + '-' + timetag + '-' + ranktype + '-' + \
            psize + '.png'] + commonargs + ['-l', '0'] + sizeargs[psize] + \
            timeargs + data_defs + plot_defs
    rrdtool.graph( *pargs )

def doplot_wait(glist, dbtype, psize, timetag, sizeargs, timeargs, ranktype):

    grouplist = glist[:]
    defs = list()
    plots = list()

    data_defs = list()
    plot_defs = list()

    skipgroup = 0
    if dbtype == 'waittime':
        if 'unused'  in grouplist:
            grouplist.remove('unused')
            skipgroup += 1
        if 'offline' in grouplist:
            grouplist.remove('offline')
            skipgroup += 1

    for group in (grouplist + ['rollover','lastroll']):
        data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')

    for idx in range(len(grouplist)):
        group = grouplist[idx]
        if group == 'unused':
            acolor = '#d8d8d8'
            skipgroup += 1
        elif group == "offline":
            acolor = "#790ead"
            skipgroup += 1
        else:
            acolor = colors[idx+skipgroup]
        pdefstr = 'LINE3' ':' + group + acolor + ':' + group
        pdefstr = pdefstr + '\\n'
        plot_defs.append(pdefstr)

    plot_defs.append('LINE2:rollover#660198')
    plot_defs.append('LINE2:lastroll#000000')

    pargs = [ PLOTDIR + dbtype + '-' + timetag + '-' + ranktype + '-' + \
            psize + '.png'] + ['--slope-mode', '-o'] + commonargs + \
            sizeargs[psize] + timeargs + data_defs + plot_defs
    rrdtool.graph( *pargs )

def makeplots(prangedef):

    resolu = prangedef['avres']
    NUMGROUPS = prangedef['numgroups']
    
    # first need to find "top eight" list
    # base it on running jobs

    now=int(time.mktime(time.localtime()))
    end = (now / resolu) * resolu
    start  = end - (prangedef['avrange']) + resolu

    ### block finding 'top N' group list ###

    tgroup = dict() # structure tgroup[groupname] = total of hourly average

    running_files = glob.glob(DATADIR+'*.running.rrd')
    for db in running_files:
        group = db[len(DATADIR):db.find('.running.rrd')]
        tup = rrdtool.fetch(db,'AVERAGE','-r', repr(resolu),
                            '-s', repr(start), '-e', repr(end))
        vallist = [0] # start with zero, in case no vals returned, get zero as answer
        for tup2 in tup[2]:
            val = tup2[0]
            if val:
                vallist.append(val)

        # put numbers in meaningful units now.  result returned is an
        # integration over the time range, of averages over "resolu"
        # ... native resolution is in minutes, so multiplying by
        # (resolu / 60) puts the answer in core-minutes; dividing by
        # the number of minutes in the range gives the average number
        # of cores occupied, over the range

        if group == "total":
            totval = sum(vallist) * (resolu / 60) / ( (end-start) / 60. )
        else:
            tgroup[group] = sum(vallist) * (resolu / 60) / ( (end-start) / 60. )

    pgroup = dict()
    for g in tgroup.keys():
        pgroup[g] = 100*tgroup[g]/totval
    
    ### start block 'top N' and 'next N' group list ###
        
    groups_sorted = sorted(tgroup, key=tgroup.get, reverse=False) # increasing useage towards end of list

    topgroups  = groups_sorted[  -NUMGROUPS:          ]
    bottgroups = groups_sorted[-2*NUMGROUPS:-NUMGROUPS]

    for glist in [topgroups, bottgroups]:
        if 'unused' in glist:
            glist.remove('unused')
            glist.append('unused')  # makes it always appear at top of plot (except see offline)
        if 'offline' in glist:
            glist.remove('offline')
            glist.append('offline')  # makes it always appear at top of plot

    if opts.rankonly and prangedef['timetag'] != 'hr' :
        print "Ranked average running jobs over period", prangedef['timetag']
        print "%10s   %12d" % ('total', totval)
        rank = 0
        for g in reversed(groups_sorted):
            if tgroup[g] > 0:
                rank += 1
                print "%2d. %10s   %9.4f  %9.4f%%" % (rank, g, tgroup[g], pgroup[g])
        return

    ### end block 'top N' group list ###

    ### block generating plots ###

   
    for dbtype in ['queued', 'running', 'waittime']:
        
        ### this is a bit of a hack : we only want pgroups for when it's
        ### a 'running' database and we don't want it for timetag hour.
        ### fix this up here

        if dbtype == 'running' and prangedef['timetag'] != 'hr':
            percents = pgroup
        else:
            percents = None
        for grps in [ ('top', topgroups), ('bottom', bottgroups) ]:
            for psize in ['small', 'large']:
                if dbtype == 'waittime':
                    doplot_wait(grps[1], dbtype, psize, prangedef['timetag'],
                                prangedef['sizeargs'],
                                prangedef['timeargs'], grps[0]
                                )
                else:
                    doplot(grps[1], dbtype, psize, prangedef['timetag'],
                           prangedef['sizeargs'],
                           prangedef['timeargs'],
                           percents, grps[0]
                           )

if opts.rankonly:
    from signal import signal, SIGPIPE, SIG_DFL
    signal(SIGPIPE,SIG_DFL)

for k in ['hr', 'day', 'week', 'month', 'year', 'alltime']:  #  plotrangedef.keys():
    makeplots(plotrangedef[k])
    
import sys

sys.stdout.flush()
sys.stdout.close()

sys.stderr.flush()
sys.stderr.close()

sys.exit(0)
1	templon	2355	#! /usr/bin/env python
2	templon	2377	# $Id$
3			# Source: $URL$
4			# J. A. Templon, NIKHEF/PDP 2011
5	templon	2355
6	templon	2556	import optparse
7
8			p = optparse.OptionParser(description="Program to make rrdtool plots " + \
9			"of job running jobs by unix group")
10
11			# p.add_option("-r",action="store",dest="minsize",default='0',help="minimum size of dirs considered; can use suffixes k,M,G for multiples of 1000**{1,2,3} bytes")
12			# p.add_option("--qdel",action="store_true",dest="deljobs",help="delete jobs for which TMPDIR is larger than MINSIZE",default=False)
13
14			p.add_option("--rank-only",action="store_true",dest="rankonly",
15			help="don't plot, just print ranking of groups",default=False)
16
17			debug = 0
18
19			opts, args = p.parse_args()
20
21	templon	2357	import os
22
23			DATADIR=os.environ['HOME'] + '/ndpfdata/'
24	templon	3113	PLOTDIR='/www/grid/stats/groupviews/ndpf/'
25	templon	3061
26	templon	2358	# 8 class qualitative paired color scheme
27
28			colors = [ "#A6CEE3", "#1F77B4", "#B2DF8A", "#33A02C",
29			"#FB9A99", "#E31A1C", "#FDBF6F", "#FF7F00" ]
30
31			colors.reverse()
32
33	templon	2556	# for reference : ranges of the RRAs
34			# base step size 60 sec
35			# step 1 : 60 sec x 1600 points : 1600 min : 26,67 hr : 1.11 days
36			# step 2 : 120 sec x 1200 points : 2400 min : 40 hr : 1,67 days
37			# step 10 : 600 sec x 1800 points : 18 000 min : 300 hr : 12,5 days
38			# step 30 : 1800 sec x 2500 points : 75 000 min : 1250 hr : 52.08 days
39			# step 120 : 7200 sec x 1000 points : 120 000 min : 2000 hr : 83.33 days
40			# step 480 : 28800 sec x 1000 points : 480 000 min : 8000 hr : 333.33 days
41			# step 1440 : 86400 sec x 3650 points : 3650 days : 10 years
42	templon	2355
43	templon	2556	# resolutions of RRAs
44
45			# 1 : 60 sec : 1 min
46			# 2 : 120 sec : 2 min
47			# 3 : 600 sec : 10 min
48			# 4 : 1 800 sec : 30 min
49			# 5 : 7 200 sec : 120 min : 2 hr
50			# 6 : 28 800 sec : 480 min : 8 hr
51			# 7 : 86 400 sec : 1440 min : 24 hr : 1 day
52
53	templon	3268	NUMQS = 33 # number of quarters for "alltime" plot
54	templon	3151
55	templon	2398	plotrangedef = {
56	templon	2556	'hr' : { 'timeargs' : [ '-s', 'n-200min', '-e', 'n' ],
57	templon	2398	'timetag' : 'hr',
58	templon	3268	'avrange' : 3*3600,
59	templon	2398	'avres' : 60,
60	templon	2556	'sizeargs' : { 'small' : [ '--width', '200', '--height', '125',
61			'--x-grid',
62			'MINUTE:20:HOUR:1:HOUR:1:0:%H:00'
63			],
64			'large' : [ '--width', '800', '--height', '500' ]
65	templon	2398	},
66	templon	3268	'numgroups' : 4,
67	templon	2398	},
68	templon	2556	'day' : { 'timeargs' : [ '-s', 'n-2000min', '-e', 'n' ],
69	templon	2398	'timetag' : 'day',
70			'avrange' : 24*3600,
71			'avres' : 60,
72	templon	2556	'sizeargs' : { 'small' : [ '--width', '200', '--height', '125',
73			'--x-grid',
74			'HOUR:6:DAY:1:HOUR:12:0:%a %H:00'
75			],
76			'large' : [ '--width', '1000', '--height', '625' ]
77	templon	2398	},
78	templon	3268	'numgroups' : 5,
79	templon	2399	},
80	templon	2556	'week' : { 'timeargs' : [ '-s', 'n-288hr', '-e', 'n' ],
81	templon	2399	'timetag' : 'week',
82			'avrange' : 7243600,
83	templon	2556	'avres' : 600,
84			'sizeargs' : { 'small' : [ '--width', '576', '--height', '125' ],
85			'large' : [ '--width', '1728', '--height', '375',
86			'-n', 'DEFAULT:16:']
87	templon	2399	},
88	templon	2821	'numgroups' : 6,
89	templon	2399	},
90
91	templon	2556	# note the construction "repr(576*n)" here --- this is because the plot is
92			# (a multiple of) 576 pixels, and 120 min is one of the RRAs, so choosing
93			# a lower limit of 576*120 gives us a plot with one pixel per RRA bin.
94
95			'month' : { 'timeargs' : [ '-s', 'n-'+repr(576*120)+'min', '-e', 'n'],
96	templon	2399	'timetag' : 'month',
97			'avrange' : 31243600,
98	templon	2556	'avres' : 1800,
99			'sizeargs' : { 'small' : [ '--width', '576', '--height', '105'],
100			'large' : [ '--width', '2304', '--height', '420',
101			'-n', 'DEFAULT:18:',
102			'--x-grid',
103			'HOUR:12:DAY:1:DAY:3:86400:%d-%b'
104			]
105	templon	2399	},
106	templon	3268	'numgroups' : 7,
107	templon	2556	},
108
109	templon	2615	'year' : { 'timeargs' : [ '-s', 'n-'+repr(576*1440)+'min', '-e', 'n'],
110	templon	2556	'timetag' : 'year',
111			'avrange' : 365243600,
112			'avres' : 86400,
113			'sizeargs' : { 'small' : [ '--width', '576', '--height', '105'],
114			'large' : [ '--width', '2304', '--height', '420']
115			},
116	templon	2821	'numgroups' : 8,
117	templon	2556	},
118	templon	2616
119	templon	3151	# adjusting for nice plot ... set NUMQS above; range is NUMQS/4 years
120	templon	2661
121	templon	3151	'alltime' : { 'timeargs' : [ '-s', 'n-'+repr(NUMQS3651440/4)+'min', '-e', 'n'],
122	templon	2618	'timetag' : 'alltime',
123	templon	3151	'avrange' : NUMQS36524*3600/4,
124	templon	2618	'avres' : 86400,
125			'sizeargs' : { 'small' : [ '--width', '576', '--height', '105',
126			'--x-grid',
127			'MONTH:3:YEAR:1:YEAR:1:31536000:%Y'
128			],
129			'large' : [ '--width', '2304', '--height', '420',
130			'--x-grid',
131			'MONTH:1:YEAR:1:MONTH:3:2592000:%b-%Y'
132			]
133			},
134	templon	2821	'numgroups' : 8,
135	templon	2616	},
136	templon	2356
137	templon	2618
138			}
139
140	templon	2556	commonargs = ['--imgformat', 'PNG',
141	templon	2399	'--legend-position=east', '--legend-direction=bottomup']
142
143	templon	2398	import rrdtool
144	templon	2357	import time
145	templon	2398	import glob
146	templon	2357
147	templon	2398	### function definitions
148	templon	2362
149	templon	2663	def doplot(glist, dbtype, psize, timetag, sizeargs, timeargs, pcents, ranktype):
150	templon	2362
151	templon	2624	grouplist = glist[:]
152	templon	2362	defs = list()
153			plots = list()
154	templon	2358
155	templon	2362	data_defs = list()
156			plot_defs = list()
157
158	templon	2677	gcolors = dict()
159
160	templon	3131	skipgroup = 0
161	templon	2677	if ranktype == 'bottom':
162			if 'unused' in grouplist:
163			grouplist.remove('unused')
164			grouplist.insert(0,'unused')
165			if 'offline' in grouplist:
166			grouplist.remove('offline')
167			grouplist.insert(0,'offline')
168	templon	3131	else:
169			if 'unused' in grouplist:
170			skipgroup += 1
171			if 'offline' in grouplist:
172			skipgroup += 1
173
174			for idx in range(len(grouplist)):
175			thisgrp = grouplist[idx]
176			if thisgrp == 'unused':
177			gcolors[thisgrp] = '#d8d8d8'
178			elif thisgrp == "offline":
179			gcolors[thisgrp] = "#790ead"
180			else:
181			gcolors[thisgrp] = colors[idx+skipgroup]
182
183	templon	2624	if dbtype == 'queued':
184			if 'unused' in grouplist: grouplist.remove('unused')
185			if 'offline' in grouplist: grouplist.remove('offline')
186
187	templon	2663	if ranktype == 'top':
188			for group in (grouplist + ['total']):
189			data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')
190			otherstring = 'CDEF:other=total,'
191			for group in grouplist:
192			otherstring += group + ','
193			otherstring += (len(grouplist)-1) * '+,' + '-'
194	templon	2362
195	templon	3113	# print otherstring
196	templon	2663	data_defs.append(otherstring)
197			elif ranktype == 'bottom':
198			for group in grouplist:
199			data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')
200			else:
201			print 'Unknown ranktype detected:', ranktype
202			sys.exit(2)
203	templon	2623
204	templon	2617	sumshown = 0
205	templon	2362	for idx in range(len(grouplist)):
206			group = grouplist[idx]
207	templon	3131	acolor = gcolors[group]
208	templon	2677	pdefstr = 'AREA' ':' + group + acolor + ':' + "%8s" % (group)
209	templon	2556	if pcents:
210			pdefstr = pdefstr + ' (' + "%4.1f" % (pcents[group]) + ')'
211	templon	2617	sumshown += float(pcents[group])
212	templon	2380	pdefstr = pdefstr + '\\n'
213	templon	2623	pdefstr = pdefstr + ':STACK'
214	templon	2362	plot_defs.append(pdefstr)
215
216	templon	2663	if ranktype == 'top':
217	templon	2677	pdefstr = 'AREA' ':' + 'other' + '#794044' + ':' + ' other'
218	templon	2663	if pcents:
219			pdefstr = pdefstr + ' (' + "%4.1f" % (100 - sumshown) + ')'
220			pdefstr = pdefstr + '\\n'
221			plot_defs.insert(0,pdefstr)
222			plot_defs.append("LINE:total#000000") # :total")
223	templon	2362
224	templon	2663	pargs = [ PLOTDIR + dbtype + '-' + timetag + '-' + ranktype + '-' + \
225			psize + '.png'] + commonargs + ['-l', '0'] + sizeargs[psize] + \
226			timeargs + data_defs + plot_defs
227	templon	2362	rrdtool.graph( *pargs )
228
229	templon	2663	def doplot_wait(glist, dbtype, psize, timetag, sizeargs, timeargs, ranktype):
230	templon	2398
231	templon	2624	grouplist = glist[:]
232	templon	2365	defs = list()
233			plots = list()
234
235			data_defs = list()
236			plot_defs = list()
237
238	templon	3131	skipgroup = 0
239	templon	2624	if dbtype == 'waittime':
240	templon	3131	if 'unused' in grouplist:
241			grouplist.remove('unused')
242			skipgroup += 1
243			if 'offline' in grouplist:
244			grouplist.remove('offline')
245			skipgroup += 1
246	templon	2624
247	templon	2556	for group in (grouplist + ['rollover','lastroll']):
248	templon	2365	data_defs.append('DEF:'+group+'='+DATADIR+group+'.'+dbtype+'.rrd:'+dbtype+':AVERAGE')
249
250			for idx in range(len(grouplist)):
251			group = grouplist[idx]
252	templon	2623	if group == 'unused':
253	templon	2734	acolor = '#d8d8d8'
254	templon	3131	skipgroup += 1
255	templon	2623	elif group == "offline":
256			acolor = "#790ead"
257	templon	3131	skipgroup += 1
258	templon	2623	else:
259	templon	3131	acolor = colors[idx+skipgroup]
260	templon	2623	pdefstr = 'LINE3' ':' + group + acolor + ':' + group
261	templon	2380	pdefstr = pdefstr + '\\n'
262	templon	2365	plot_defs.append(pdefstr)
263
264	templon	2556	plot_defs.append('LINE2:rollover#660198')
265			plot_defs.append('LINE2:lastroll#000000')
266	templon	2365
267	templon	2663	pargs = [ PLOTDIR + dbtype + '-' + timetag + '-' + ranktype + '-' + \
268			psize + '.png'] + ['--slope-mode', '-o'] + commonargs + \
269			sizeargs[psize] + timeargs + data_defs + plot_defs
270	templon	2365	rrdtool.graph( *pargs )
271
272	templon	2398	def makeplots(prangedef):
273
274			resolu = prangedef['avres']
275	templon	2821	NUMGROUPS = prangedef['numgroups']
276	templon	2398
277			# first need to find "top eight" list
278			# base it on running jobs
279
280			now=int(time.mktime(time.localtime()))
281			end = (now / resolu) * resolu
282			start = end - (prangedef['avrange']) + resolu
283
284			### block finding 'top N' group list ###
285
286			tgroup = dict() # structure tgroup[groupname] = total of hourly average
287
288			running_files = glob.glob(DATADIR+'*.running.rrd')
289			for db in running_files:
290			group = db[len(DATADIR):db.find('.running.rrd')]
291			tup = rrdtool.fetch(db,'AVERAGE','-r', repr(resolu),
292			'-s', repr(start), '-e', repr(end))
293			vallist = [0] # start with zero, in case no vals returned, get zero as answer
294			for tup2 in tup[2]:
295			val = tup2[0]
296			if val:
297			vallist.append(val)
298	templon	2579
299	templon	2663	# put numbers in meaningful units now. result returned is an
300			# integration over the time range, of averages over "resolu"
301			# ... native resolution is in minutes, so multiplying by
302			# (resolu / 60) puts the answer in core-minutes; dividing by
303			# the number of minutes in the range gives the average number
304			# of cores occupied, over the range
305	templon	2579
306	templon	2556	if group == "total":
307	templon	2579	totval = sum(vallist) * (resolu / 60) / ( (end-start) / 60. )
308	templon	2556	else:
309	templon	2579	tgroup[group] = sum(vallist) * (resolu / 60) / ( (end-start) / 60. )
310	templon	2398
311	templon	2556	pgroup = dict()
312			for g in tgroup.keys():
313			pgroup[g] = 100*tgroup[g]/totval
314
315	templon	2663	### start block 'top N' and 'next N' group list ###
316
317	templon	2661	groups_sorted = sorted(tgroup, key=tgroup.get, reverse=False) # increasing useage towards end of list
318
319	templon	2663	topgroups = groups_sorted[ -NUMGROUPS: ]
320			bottgroups = groups_sorted[-2*NUMGROUPS:-NUMGROUPS]
321	templon	2398
322	templon	2663	for glist in [topgroups, bottgroups]:
323			if 'unused' in glist:
324			glist.remove('unused')
325			glist.append('unused') # makes it always appear at top of plot (except see offline)
326			if 'offline' in glist:
327			glist.remove('offline')
328			glist.append('offline') # makes it always appear at top of plot
329
330	templon	2579	if opts.rankonly and prangedef['timetag'] != 'hr' :
331			print "Ranked average running jobs over period", prangedef['timetag']
332			print "%10s %12d" % ('total', totval)
333			rank = 0
334	templon	2556	for g in reversed(groups_sorted):
335			if tgroup[g] > 0:
336	templon	2579	rank += 1
337			print "%2d. %10s %9.4f %9.4f%%" % (rank, g, tgroup[g], pgroup[g])
338	templon	2556	return
339
340	templon	2398	### end block 'top N' group list ###
341
342			### block generating plots ###
343
344	templon	2556
345	templon	2398	for dbtype in ['queued', 'running', 'waittime']:
346	templon	2556
347			### this is a bit of a hack : we only want pgroups for when it's
348			### a 'running' database and we don't want it for timetag hour.
349			### fix this up here
350
351			if dbtype == 'running' and prangedef['timetag'] != 'hr':
352			percents = pgroup
353			else:
354			percents = None
355	templon	2663	for grps in [ ('top', topgroups), ('bottom', bottgroups) ]:
356			for psize in ['small', 'large']:
357			if dbtype == 'waittime':
358			doplot_wait(grps[1], dbtype, psize, prangedef['timetag'],
359			prangedef['sizeargs'],
360			prangedef['timeargs'], grps[0]
361			)
362			else:
363			doplot(grps[1], dbtype, psize, prangedef['timetag'],
364			prangedef['sizeargs'],
365			prangedef['timeargs'],
366			percents, grps[0]
367			)
368	templon	2362
369	templon	3239	if opts.rankonly:
370			from signal import signal, SIGPIPE, SIG_DFL
371			signal(SIGPIPE,SIG_DFL)
372
373	templon	2616	for k in ['hr', 'day', 'week', 'month', 'year', 'alltime']: # plotrangedef.keys():
374	templon	2399	makeplots(plotrangedef[k])
375
376	templon	2357	import sys
377	templon	3239
378			sys.stdout.flush()
379			sys.stdout.close()
380
381			sys.stderr.flush()
382			sys.stderr.close()
383
384	templon	2357	sys.exit(0)
grid.support@nikhef.nl	ViewVC Help
Powered by ViewVC 1.1.28