#!/usr/bin/env python # $Id$ # Source: $URL$ # 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. # constants to be changed each year pledges = { # pledges in HS06 units 'alice' : 1844, 'atlas' : 17580, 'lhcb' : 8827 } total_cap = 39671 import optparse usage = "usage: %prog -p DBpwd start-date [end-date]\n" + \ "date format yyyy-mm-dd; default end date is 3 months after start date" p = optparse.OptionParser(description="Program to generate a CSV dump of computing usage " + \ " by WLCG over a specified period",usage=usage) # 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("Format: yyyy-mm-dd\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'] } 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([' ',' ']) writer.writerow(["Pledges and capacties (HS06)"]) writer.writerow(["Group", "power"]) writer.writerow(["ATLAS", pledges["atlas"]]) writer.writerow(["LHCb", pledges["lhcb"] ]) writer.writerow(["ALICE", pledges["alice"]]) writer.writerow(["farmcap", total_cap])