nl.nikhef.ndpf.3maand/trunk/3maand.py

#!/usr/bin/env python
# $Id: ndpf-gv-mkplots 2429 2011-09-09 19:27:11Z templon $
# Source: $URL: svn+ssh://svn@ndpfsvn.nikhef.nl/repos/pdpsoft/nl.nikhef.ndpf.groupviews/branches/RB-2.1.1/ndpf-gv-mkplots $
# J. A. Templon, NIKHEF/PDP 2011

# script to do yearly WLCG accounting and format it like I want.
# note : sorting logic is based on a by-month grouping; if you change that,
#        you'll need to change some of the python code.

import optparse

p = optparse.OptionParser(description="Program to generate a CSV dump of computing usage " + \
                          " by WLCG over a specified period")

# dbpassword is mandatory
# start date is mandatory
# end date is optional; proving nothing means end date is 3 months after start
# end date can be provided in format yyyy-mm-dd or as +3m (for 3 months later than start)

p.add_option("-p",action="store",dest="dbpassw",default=None,
             help="password for NDPF accounting DB")
debug = 0

opts, args = p.parse_args()

import sys
if not opts.dbpassw :
    sys.stderr.write("Error: password to NDPF accounting DB must be provided with -p\n")
    sys.stderr.write("Bailing out.\n")
    sys.exit(1)
    
if len(args) < 1:
    sys.stderr.write("Error: no date argument detected.  A start date must be provided.\n")
    sys.stderr.write("Bailing out.\n")
    sys.exit(1)
    
start_ascii = args[0]
import datetime
def parsedate(s):
    try:
        farr=s.split("-")
        iarr = list()
        for f in farr:
            iarr.append(int(f))
        return datetime.date(iarr[0],iarr[1],iarr[2])
    except:
        sys.stderr.write("Error parsing date string " + s + "\n")
        raise

if len(args) > 1:
    end_ascii = args[1]
else:
    end_ascii = None

SDATE = parsedate(start_ascii)  # starting date; script logic assumes start of month
if not end_ascii:
    nmon = 3
elif end_ascii[:1] == '+':
    if end_ascii[-1:] != 'm':
        sys.stderr.write("error in end date string.  accepted formats are"+
                         " 2011-05-27 or +3m\n")
        sys.exit(1)
    nmon = int(end_ascii[1:end_ascii.index('m')])
else:
    nmon = None

if nmon:
    eyear = SDATE.year
    emonth = SDATE.month + nmon
    if emonth > 12:
        emonth = emonth - 12
        eyear = eyear + 1
    EDATE = datetime.date(eyear,emonth,SDATE.day)
else:
    EDATE = parsedate(end_ascii)  # starting date; script logic assumes start of month

print "generating data for jobs between", SDATE, "and", EDATE
vos = ['alice', 'atlas', 'lhcb']
groups = {
    'alice': ['palice', 'alicesgm'],
    'atlas': ['atlb', 'atlaspil', 'patlas', 'atlsgm'],
    'lhcb' : ['lhcb', 'lhcbpil', 'lhcbprd', 'lhcbsgm']
    }

pledges = {          # pledges in HS06 units
    'alice' : 1844,
    'atlas' : 17580,
    'lhcb'  : 8827
    }

ACCBASECMD = ['/usr/local/bin/accuse']
ACCBASEARG = ('--byendtime -dbpasswd ' + opts.dbpassw + ' -f lcg -m').split()

import subprocess

perlout=dict()

# gather raw accuse output

for vo in vos:
    groupargs = []
    for g in groups[vo]:
        groupargs += ["-g",g]
    args = ACCBASECMD + ACCBASEARG + ["-s", SDATE.isoformat(), "-e", EDATE.isoformat()] + groupargs
    perlout[vo] = subprocess.Popen(args, stdout=subprocess.PIPE).communicate()[0]

def hms2dec(str):
    h,m,s = str.split(':')
    return float(h) + (float(m) + float(s)/60.)/60.

# parse output
# results in data structure like this

# data = parsed['alice']['2010-02']['hs06days']

import re
mpatt = re.compile(r'20[012][0-9]-[01][0-9]')
parsed = dict()
for vo in vos:
    parsed[vo] = dict()
    lines = perlout[vo].split('\n')
    for line in lines:
        if mpatt.match(line) or line.find('Summed') == 0:
            fields = line.split()
            cpu = hms2dec(fields[1])
            wall = hms2dec(fields[2])
            hs06days = float(fields[4]) * 4
            njobs = int(fields[5])
            parsed[vo][fields[0]] = { 'cpu' : cpu, 'wall': wall,
                                      'hs06days': hs06days, 'njobs' : njobs }

# output generation ... big damn csv file

import csv
writer = csv.writer(open('tmp.csv', 'wb'), delimiter=',',
                    quotechar='|', quoting=csv.QUOTE_MINIMAL)

ONEDAY=datetime.timedelta(days=1)

# per-VO segment

for vo in vos:
    writer.writerow(["Data for",vo])
    writer.writerow(["Month", "hs06days.used", "cpu/wall", "pledged", "jobs", "days"])
    
    monthstart = SDATE

    while monthstart < EDATE:

        if monthstart.month < 12 :
            startnextmonth = monthstart.replace(month=monthstart.month+1)
        else:
            d1 = monthstart.replace(month=1)
            startnextmonth = d1.replace(year=monthstart.year+1)
        monthend =  startnextmonth - ONEDAY
        if monthend > EDATE:
            monthend = EDATE
        ndays = (monthend - monthstart + ONEDAY).days
        monthkey = monthstart.isoformat()[:7]
        if monthkey not in parsed[vo].keys():
            parsed[vo][monthkey] = {
                'hs06days' : 0, 'cpu' : 0.1, 'wall' : 0.1, 'njobs' : 0
                }
        td = parsed[vo][monthkey]
        writer.writerow([monthkey,
                         td['hs06days'],
                         td['cpu']/td['wall'],
                         ndays * pledges[vo],
                         td['njobs'],
                         ndays
                         ])
        monthstart = monthend + ONEDAY

    writer.writerow([' ',' '])

# usage plots

writer.writerow(["hs06 days used per VO"])
writer.writerow(["Month", "lhcb.used", "atlas.used", "alice.used"])
monthstart = SDATE
while monthstart < EDATE:
    if monthstart.month < 12 :
        startnextmonth = monthstart.replace(month=monthstart.month+1)
    else:
        d1 = monthstart.replace(month=1)
        startnextmonth = d1.replace(year=monthstart.year+1)
    monthend =  startnextmonth - ONEDAY
    if monthend > EDATE:
        monthend = EDATE
    ndays = (monthend - monthstart + ONEDAY).days
    monthkey = monthstart.isoformat()[:7]
    td = parsed[vo][monthkey]
    writer.writerow([monthkey,
                     parsed['lhcb'][monthkey]['hs06days'],
                     parsed['atlas'][monthkey]['hs06days'],
                     parsed['alice'][monthkey]['hs06days']
                     ])
    monthstart = monthend + ONEDAY

writer.writerow([' ',' '])

# pledge fraction plots

writer.writerow(["pledge fraction used per VO"])
writer.writerow(["Month", "lhcb.frac", "atlas.frac", "alice.frac"])
monthstart = SDATE
while monthstart < EDATE:
    if monthstart.month < 12 :
        startnextmonth = monthstart.replace(month=monthstart.month+1)
    else:
        d1 = monthstart.replace(month=1)
        startnextmonth = d1.replace(year=monthstart.year+1)
    monthend =  startnextmonth - ONEDAY
    if monthend > EDATE:
        monthend = EDATE
    ndays = (monthend - monthstart + ONEDAY).days
    monthkey = monthstart.isoformat()[:7]
    td = parsed[vo][monthkey]
    writer.writerow([monthkey,
                     parsed['lhcb'][monthkey]['hs06days']/(pledges['lhcb']*ndays),
                     parsed['atlas'][monthkey]['hs06days']/(pledges['atlas']*ndays),
                     parsed['alice'][monthkey]['hs06days']/(pledges['alice']*ndays)
                     ])
    monthstart = monthend + ONEDAY

writer.writerow([' ',' '])

# cpu eff plots

writer.writerow(["ratio cpu to wall time used (eff) per VO"])
writer.writerow(["Month", "lhcb.eff", "atlas.eff", "alice.eff"])
monthstart = SDATE
while monthstart < EDATE:
    if monthstart.month < 12 :
        startnextmonth = monthstart.replace(month=monthstart.month+1)
    else:
        d1 = monthstart.replace(month=1)
        startnextmonth = d1.replace(year=monthstart.year+1)
    monthend =  startnextmonth - ONEDAY
    if monthend > EDATE:
        monthend = EDATE
    ndays = (monthend - monthstart + ONEDAY).days
    monthkey = monthstart.isoformat()[:7]
    td = parsed[vo][monthkey]
    writer.writerow([monthkey,
                     parsed['lhcb'][monthkey]['cpu']/parsed['lhcb'][monthkey]['wall'],
                     parsed['atlas'][monthkey]['cpu']/parsed['atlas'][monthkey]['wall'],
                     parsed['alice'][monthkey]['cpu']/parsed['alice'][monthkey]['wall'],
                     ])
    monthstart = monthend + ONEDAY

writer.writerow([' ',' '])
1	#!/usr/bin/env python
2	# $Id: ndpf-gv-mkplots 2429 2011-09-09 19:27:11Z templon $
3	# Source: $URL: svn+ssh://svn@ndpfsvn.nikhef.nl/repos/pdpsoft/nl.nikhef.ndpf.groupviews/branches/RB-2.1.1/ndpf-gv-mkplots $
4	# J. A. Templon, NIKHEF/PDP 2011
5
6	# script to do yearly WLCG accounting and format it like I want.
7	# note : sorting logic is based on a by-month grouping; if you change that,
8	# you'll need to change some of the python code.
9
10	import optparse
11
12	p = optparse.OptionParser(description="Program to generate a CSV dump of computing usage " + \
13	" by WLCG over a specified period")
14
15	# dbpassword is mandatory
16	# start date is mandatory
17	# end date is optional; proving nothing means end date is 3 months after start
18	# end date can be provided in format yyyy-mm-dd or as +3m (for 3 months later than start)
19
20	p.add_option("-p",action="store",dest="dbpassw",default=None,
21	help="password for NDPF accounting DB")
22	debug = 0
23
24	opts, args = p.parse_args()
25
26	import sys
27	if not opts.dbpassw :
28	sys.stderr.write("Error: password to NDPF accounting DB must be provided with -p\n")
29	sys.stderr.write("Bailing out.\n")
30	sys.exit(1)
31
32	if len(args) < 1:
33	sys.stderr.write("Error: no date argument detected. A start date must be provided.\n")
34	sys.stderr.write("Bailing out.\n")
35	sys.exit(1)
36
37	start_ascii = args[0]
38	import datetime
39	def parsedate(s):
40	try:
41	farr=s.split("-")
42	iarr = list()
43	for f in farr:
44	iarr.append(int(f))
45	return datetime.date(iarr[0],iarr[1],iarr[2])
46	except:
47	sys.stderr.write("Error parsing date string " + s + "\n")
48	raise
49
50	if len(args) > 1:
51	end_ascii = args[1]
52	else:
53	end_ascii = None
54
55	SDATE = parsedate(start_ascii) # starting date; script logic assumes start of month
56	if not end_ascii:
57	nmon = 3
58	elif end_ascii[:1] == '+':
59	if end_ascii[-1:] != 'm':
60	sys.stderr.write("error in end date string. accepted formats are"+
61	" 2011-05-27 or +3m\n")
62	sys.exit(1)
63	nmon = int(end_ascii[1:end_ascii.index('m')])
64	else:
65	nmon = None
66
67	if nmon:
68	eyear = SDATE.year
69	emonth = SDATE.month + nmon
70	if emonth > 12:
71	emonth = emonth - 12
72	eyear = eyear + 1
73	EDATE = datetime.date(eyear,emonth,SDATE.day)
74	else:
75	EDATE = parsedate(end_ascii) # starting date; script logic assumes start of month
76
77	print "generating data for jobs between", SDATE, "and", EDATE
78	vos = ['alice', 'atlas', 'lhcb']
79	groups = {
80	'alice': ['palice', 'alicesgm'],
81	'atlas': ['atlb', 'atlaspil', 'patlas', 'atlsgm'],
82	'lhcb' : ['lhcb', 'lhcbpil', 'lhcbprd', 'lhcbsgm']
83	}
84
85	pledges = { # pledges in HS06 units
86	'alice' : 1844,
87	'atlas' : 17580,
88	'lhcb' : 8827
89	}
90
91	ACCBASECMD = ['/usr/local/bin/accuse']
92	ACCBASEARG = ('--byendtime -dbpasswd ' + opts.dbpassw + ' -f lcg -m').split()
93
94	import subprocess
95
96	perlout=dict()
97
98	# gather raw accuse output
99
100	for vo in vos:
101	groupargs = []
102	for g in groups[vo]:
103	groupargs += ["-g",g]
104	args = ACCBASECMD + ACCBASEARG + ["-s", SDATE.isoformat(), "-e", EDATE.isoformat()] + groupargs
105	perlout[vo] = subprocess.Popen(args, stdout=subprocess.PIPE).communicate()[0]
106
107	def hms2dec(str):
108	h,m,s = str.split(':')
109	return float(h) + (float(m) + float(s)/60.)/60.
110
111	# parse output
112	# results in data structure like this
113
114	# data = parsed['alice']['2010-02']['hs06days']
115
116	import re
117	mpatt = re.compile(r'20[012][0-9]-[01][0-9]')
118	parsed = dict()
119	for vo in vos:
120	parsed[vo] = dict()
121	lines = perlout[vo].split('\n')
122	for line in lines:
123	if mpatt.match(line) or line.find('Summed') == 0:
124	fields = line.split()
125	cpu = hms2dec(fields[1])
126	wall = hms2dec(fields[2])
127	hs06days = float(fields[4]) * 4
128	njobs = int(fields[5])
129	parsed[vo][fields[0]] = { 'cpu' : cpu, 'wall': wall,
130	'hs06days': hs06days, 'njobs' : njobs }
131
132	# output generation ... big damn csv file
133
134	import csv
135	writer = csv.writer(open('tmp.csv', 'wb'), delimiter=',',
136	quotechar='\|', quoting=csv.QUOTE_MINIMAL)
137
138	ONEDAY=datetime.timedelta(days=1)
139
140	# per-VO segment
141
142	for vo in vos:
143	writer.writerow(["Data for",vo])
144	writer.writerow(["Month", "hs06days.used", "cpu/wall", "pledged", "jobs", "days"])
145
146	monthstart = SDATE
147
148	while monthstart < EDATE:
149
150	if monthstart.month < 12 :
151	startnextmonth = monthstart.replace(month=monthstart.month+1)
152	else:
153	d1 = monthstart.replace(month=1)
154	startnextmonth = d1.replace(year=monthstart.year+1)
155	monthend = startnextmonth - ONEDAY
156	if monthend > EDATE:
157	monthend = EDATE
158	ndays = (monthend - monthstart + ONEDAY).days
159	monthkey = monthstart.isoformat()[:7]
160	if monthkey not in parsed[vo].keys():
161	parsed[vo][monthkey] = {
162	'hs06days' : 0, 'cpu' : 0.1, 'wall' : 0.1, 'njobs' : 0
163	}
164	td = parsed[vo][monthkey]
165	writer.writerow([monthkey,
166	td['hs06days'],
167	td['cpu']/td['wall'],
168	ndays * pledges[vo],
169	td['njobs'],
170	ndays
171	])
172	monthstart = monthend + ONEDAY
173
174	writer.writerow([' ',' '])
175
176	# usage plots
177
178	writer.writerow(["hs06 days used per VO"])
179	writer.writerow(["Month", "lhcb.used", "atlas.used", "alice.used"])
180	monthstart = SDATE
181	while monthstart < EDATE:
182	if monthstart.month < 12 :
183	startnextmonth = monthstart.replace(month=monthstart.month+1)
184	else:
185	d1 = monthstart.replace(month=1)
186	startnextmonth = d1.replace(year=monthstart.year+1)
187	monthend = startnextmonth - ONEDAY
188	if monthend > EDATE:
189	monthend = EDATE
190	ndays = (monthend - monthstart + ONEDAY).days
191	monthkey = monthstart.isoformat()[:7]
192	td = parsed[vo][monthkey]
193	writer.writerow([monthkey,
194	parsed['lhcb'][monthkey]['hs06days'],
195	parsed['atlas'][monthkey]['hs06days'],
196	parsed['alice'][monthkey]['hs06days']
197	])
198	monthstart = monthend + ONEDAY
199
200	writer.writerow([' ',' '])
201
202	# pledge fraction plots
203
204	writer.writerow(["pledge fraction used per VO"])
205	writer.writerow(["Month", "lhcb.frac", "atlas.frac", "alice.frac"])
206	monthstart = SDATE
207	while monthstart < EDATE:
208	if monthstart.month < 12 :
209	startnextmonth = monthstart.replace(month=monthstart.month+1)
210	else:
211	d1 = monthstart.replace(month=1)
212	startnextmonth = d1.replace(year=monthstart.year+1)
213	monthend = startnextmonth - ONEDAY
214	if monthend > EDATE:
215	monthend = EDATE
216	ndays = (monthend - monthstart + ONEDAY).days
217	monthkey = monthstart.isoformat()[:7]
218	td = parsed[vo][monthkey]
219	writer.writerow([monthkey,
220	parsed['lhcb'][monthkey]['hs06days']/(pledges['lhcb']*ndays),
221	parsed['atlas'][monthkey]['hs06days']/(pledges['atlas']*ndays),
222	parsed['alice'][monthkey]['hs06days']/(pledges['alice']*ndays)
223	])
224	monthstart = monthend + ONEDAY
225
226	writer.writerow([' ',' '])
227
228	# cpu eff plots
229
230	writer.writerow(["ratio cpu to wall time used (eff) per VO"])
231	writer.writerow(["Month", "lhcb.eff", "atlas.eff", "alice.eff"])
232	monthstart = SDATE
233	while monthstart < EDATE:
234	if monthstart.month < 12 :
235	startnextmonth = monthstart.replace(month=monthstart.month+1)
236	else:
237	d1 = monthstart.replace(month=1)
238	startnextmonth = d1.replace(year=monthstart.year+1)
239	monthend = startnextmonth - ONEDAY
240	if monthend > EDATE:
241	monthend = EDATE
242	ndays = (monthend - monthstart + ONEDAY).days
243	monthkey = monthstart.isoformat()[:7]
244	td = parsed[vo][monthkey]
245	writer.writerow([monthkey,
246	parsed['lhcb'][monthkey]['cpu']/parsed['lhcb'][monthkey]['wall'],
247	parsed['atlas'][monthkey]['cpu']/parsed['atlas'][monthkey]['wall'],
248	parsed['alice'][monthkey]['cpu']/parsed['alice'][monthkey]['wall'],
249	])
250	monthstart = monthend + ONEDAY
251
252	writer.writerow([' ',' '])
grid.support@nikhef.nl	ViewVC Help
Powered by ViewVC 1.1.28