| 1 |
gutsche |
1.1 |
#!/usr/bin/env perl
|
| 2 |
|
|
|
| 3 |
|
|
##H This drop box agent initiates tasks on all published datasets.
|
| 4 |
|
|
##H A task is an application to run on a dataset at a particular site.
|
| 5 |
|
|
##H It keeps a record of tasks created and avoids submitting a task too
|
| 6 |
|
|
##H frequently.
|
| 7 |
|
|
##H
|
| 8 |
|
|
##H Usage:
|
| 9 |
|
|
##H TaskSource
|
| 10 |
|
|
##H -state DIRECTORY [-next NEXT] [-wait SECS] [-url URL-PUBLISHED]
|
| 11 |
|
|
##H [-ignore-sites REGEXP] [-accept-sites REGEXP]
|
| 12 |
|
|
##H [-secs-per-event N] [-max-site-queue N]
|
| 13 |
|
|
##H
|
| 14 |
|
|
##H -state agent state directory, including inbox
|
| 15 |
|
|
##H -next next agent to pass the drops to; can be given several times
|
| 16 |
|
|
##H -wait time to wait in seconds between work scans
|
| 17 |
|
|
##H -url contact string for published datasets
|
| 18 |
|
|
##H -ignore-sites
|
| 19 |
|
|
##H regular expression for sites to ignore; ignore applies before
|
| 20 |
|
|
##H accept, and by default nothing is ignored and everything is
|
| 21 |
|
|
##H accepted. applies to pubdb site names, not the host names of
|
| 22 |
|
|
##H the site.
|
| 23 |
|
|
##H -accept-sites
|
| 24 |
|
|
##H regular expression for sites to accept; ignore applies before
|
| 25 |
|
|
##H accept, and by default nothing is ignored and everything is
|
| 26 |
|
|
##H accepted. applies to pubdb site names, not the host names of
|
| 27 |
|
|
##H the site.
|
| 28 |
|
|
##H -secs-per-event
|
| 29 |
|
|
##H minimum time to allocate for a task, given a number of events
|
| 30 |
|
|
##H in the dataset to process. tasks are not created more often
|
| 31 |
|
|
##H than this.
|
| 32 |
|
|
##H -max-site-queue
|
| 33 |
|
|
##H the high water mark of currently submitted jobs to a site,
|
| 34 |
|
|
##H above which new tasks will not be created.
|
| 35 |
|
|
|
| 36 |
|
|
BEGIN {
|
| 37 |
|
|
use strict; use warnings; $^W=1;
|
| 38 |
|
|
our $me = $0; $me =~ s|.*/||;
|
| 39 |
|
|
our $home = $0; $home =~ s|/[^/]+$||; $home ||= "."; $home .= "/../PHEDEX/Toolkit/Common";
|
| 40 |
|
|
unshift(@INC, $home);
|
| 41 |
|
|
}
|
| 42 |
|
|
|
| 43 |
|
|
######################################################################
|
| 44 |
|
|
use UtilsHelp;
|
| 45 |
|
|
my %args = (WAITTIME => 600, SECS_PER_EVENT => 1., MAX_SITE_QUEUE => 10,
|
| 46 |
|
|
URL => "http://cmsdoc.cern.ch/cms/production/www/PubDB/GetPublishedCollectionInfoFromRefDB.php");
|
| 47 |
|
|
while (scalar @ARGV)
|
| 48 |
|
|
{
|
| 49 |
|
|
if ($ARGV[0] eq '-state' && scalar @ARGV > 1)
|
| 50 |
|
|
{ shift (@ARGV); $args{DROPDIR}= shift(@ARGV); }
|
| 51 |
|
|
elsif ($ARGV[0] eq '-next' && scalar @ARGV > 1)
|
| 52 |
|
|
{ shift (@ARGV); push (@{$args{NEXTDIR}}, shift(@ARGV)); }
|
| 53 |
|
|
elsif ($ARGV[0] eq '-wait' && scalar @ARGV > 1)
|
| 54 |
|
|
{ shift (@ARGV); $args{WAITTIME} = shift(@ARGV); }
|
| 55 |
|
|
elsif ($ARGV[0] eq '-ignore-sites' && scalar @ARGV > 1)
|
| 56 |
|
|
{ shift (@ARGV); $args{IGNORE_REGEXP} = shift(@ARGV); }
|
| 57 |
|
|
elsif ($ARGV[0] eq '-accept-sites' && scalar @ARGV > 1)
|
| 58 |
|
|
{ shift (@ARGV); $args{ACCEPT_REGEXP} = shift(@ARGV); }
|
| 59 |
|
|
elsif ($ARGV[0] eq '-secs-per-event' && scalar @ARGV > 1)
|
| 60 |
|
|
{ shift (@ARGV); $args{SECS_PER_EVENT} = shift(@ARGV); }
|
| 61 |
|
|
elsif ($ARGV[0] eq '-max-site-queue' && scalar @ARGV > 1)
|
| 62 |
|
|
{ shift (@ARGV); $args{MAX_SITE_QUEUE} = shift(@ARGV); }
|
| 63 |
|
|
elsif ($ARGV[0] eq '-url' && scalar @ARGV > 1)
|
| 64 |
|
|
{ shift (@ARGV); $args{URL} = shift(@ARGV); }
|
| 65 |
|
|
# Marco
|
| 66 |
|
|
elsif ($ARGV[0] eq '-dataset' && scalar @ARGV > 1)
|
| 67 |
|
|
{ shift (@ARGV); $args{DATASET} = shift(@ARGV); }
|
| 68 |
|
|
elsif ($ARGV[0] eq '-owner' && scalar @ARGV > 1)
|
| 69 |
|
|
{ shift (@ARGV); $args{OWNER} = shift(@ARGV); }
|
| 70 |
|
|
elsif ($ARGV[0] eq '-events' && scalar @ARGV > 1)
|
| 71 |
|
|
{ shift (@ARGV); $args{NEVENT} = shift(@ARGV); }
|
| 72 |
|
|
elsif ($ARGV[0] eq '-mode' && scalar @ARGV > 1)
|
| 73 |
|
|
{ shift (@ARGV); $args{MODE} = shift(@ARGV); }
|
| 74 |
|
|
elsif ($ARGV[0] eq '-scheduler' && scalar @ARGV > 1)
|
| 75 |
|
|
{ shift (@ARGV); $args{SCHEDULER} = shift(@ARGV); }
|
| 76 |
|
|
elsif ($ARGV[0] eq '-jobtype' && scalar @ARGV > 1)
|
| 77 |
|
|
{ shift (@ARGV); $args{JOBTYPE} = shift(@ARGV); }
|
| 78 |
|
|
# Marco
|
| 79 |
|
|
elsif ($ARGV[0] eq '-h')
|
| 80 |
|
|
{ &usage(); }
|
| 81 |
|
|
else
|
| 82 |
|
|
{ last; }
|
| 83 |
|
|
}
|
| 84 |
|
|
|
| 85 |
|
|
if (@ARGV || !$args{DROPDIR} || !$args{URL})
|
| 86 |
|
|
{
|
| 87 |
|
|
die "Insufficient parameters, use -h for help.\n";
|
| 88 |
|
|
}
|
| 89 |
|
|
|
| 90 |
|
|
(new TaskSource (%args))->process();
|
| 91 |
|
|
|
| 92 |
|
|
######################################################################
|
| 93 |
|
|
# Routines specific to this agent.
|
| 94 |
|
|
package TaskSource; use strict; use warnings; use base 'UtilsAgent';
|
| 95 |
|
|
use File::Path;
|
| 96 |
|
|
use UtilsCommand;
|
| 97 |
|
|
use UtilsLogging;
|
| 98 |
|
|
use UtilsTiming;
|
| 99 |
|
|
use UtilsNet;
|
| 100 |
|
|
use POSIX;
|
| 101 |
|
|
|
| 102 |
|
|
sub new
|
| 103 |
|
|
{
|
| 104 |
|
|
my $proto = shift;
|
| 105 |
|
|
my $class = ref($proto) || $proto;
|
| 106 |
|
|
my $self = $class->SUPER::new(@_);
|
| 107 |
|
|
my %params = (SECS_PER_EVENT => undef, # secs/event to delay per dataset
|
| 108 |
|
|
MAX_SITE_QUEUE => undef, # max number of jobs per site
|
| 109 |
|
|
IGNORE_REGEXP => undef, # regexp of sites to ignore
|
| 110 |
|
|
ACCEPT_REGEXP => undef, # regexp of sites to accept
|
| 111 |
|
|
DATASET => undef, # specific dataset
|
| 112 |
|
|
OWNER => undef, # specific owner
|
| 113 |
|
|
NEVENT => 500, # number of events per job
|
| 114 |
|
|
MODE => 1, # data discovery mode: (1) PudDB/RefDB, (2) DBS/DLS
|
| 115 |
|
|
JOBTYPE => "orca", # standard jobtype
|
| 116 |
|
|
SCHEDULER => "edg", # standard scheduler
|
| 117 |
|
|
URL => undef); # published dataset url
|
| 118 |
|
|
my %args = (@_);
|
| 119 |
|
|
map { $self->{$_} = defined $args{$_} ? $args{$_} : $params{$_} } keys %params;
|
| 120 |
|
|
bless $self, $class;
|
| 121 |
|
|
return $self;
|
| 122 |
|
|
}
|
| 123 |
|
|
|
| 124 |
|
|
sub init
|
| 125 |
|
|
{
|
| 126 |
|
|
my ($self) = @_;
|
| 127 |
|
|
$self->{TASKREPO} = "$self->{DROPDIR}/tasks";
|
| 128 |
|
|
-d "$self->{TASKREPO}"
|
| 129 |
|
|
|| mkdir "$self->{TASKREPO}"
|
| 130 |
|
|
|| die "$self->{TASKREPO}: cannot create directory: $!\n";
|
| 131 |
|
|
|
| 132 |
|
|
# Determine if links supports -dump-width option
|
| 133 |
|
|
$self->{LINKS_OPTS} = [];
|
| 134 |
|
|
open (LINKS_HELP, "links -help 2>/dev/null |");
|
| 135 |
|
|
if (grep (/-dump-width/, <LINKS_HELP>)) {
|
| 136 |
|
|
push(@{$self->{LINKS_OPTS}}, qw(-dump-width 300));
|
| 137 |
|
|
}
|
| 138 |
|
|
close (LINKS_HELP);
|
| 139 |
|
|
|
| 140 |
|
|
# Precode whitelist. Should really read this from somewhere...
|
| 141 |
|
|
|
| 142 |
|
|
$self->{WHITELIST} = { ASCC => "sinica.edu.tw",
|
| 143 |
|
|
NCU => "ncu.edu.tw",
|
| 144 |
|
|
FNAL => "fnal.gov",
|
| 145 |
|
|
CNAF => "webserver.infn.it",
|
| 146 |
|
|
BA => "ba.infn.it",
|
| 147 |
|
|
IN2P3=> "in2p3.fr",
|
| 148 |
|
|
PIC => "pic.es",
|
| 149 |
|
|
T2_SP=> "ciemat.es",
|
| 150 |
|
|
RAL => "ral.ac.uk",
|
| 151 |
|
|
CERN => "cern.ch",
|
| 152 |
|
|
FZK => "fzk.de",
|
| 153 |
|
|
DESY => "desy.de",
|
| 154 |
|
|
NEBR => "unl.edu",
|
| 155 |
|
|
WISC => "wisc.edu",
|
| 156 |
|
|
UFL => "ufl.edu",
|
| 157 |
|
|
PURDUE => "purdue.edu",
|
| 158 |
|
|
UCSD => "ucsd.edu",
|
| 159 |
|
|
CALT => "caltech.edu"
|
| 160 |
|
|
};
|
| 161 |
|
|
}
|
| 162 |
|
|
|
| 163 |
|
|
# Find out how many jobs are pending for each site. This is
|
| 164 |
|
|
# insensitive to the job type, and we only check once in the
|
| 165 |
|
|
# beginning to avoid favouring one dataset over another --
|
| 166 |
|
|
# once we decide to proceed for a site, we submit jobs for
|
| 167 |
|
|
# all datasets.
|
| 168 |
|
|
sub getSiteStatus
|
| 169 |
|
|
{
|
| 170 |
|
|
my ($self) = @_;
|
| 171 |
|
|
my %result = ();
|
| 172 |
|
|
my $taskrepo = $self->{TASKREPO};
|
| 173 |
|
|
foreach my $site (<$taskrepo/*/*>)
|
| 174 |
|
|
{
|
| 175 |
|
|
my ($sitename) = ($site =~ m|.*/(.*)|);
|
| 176 |
|
|
foreach my $d (<$site/*/*>)
|
| 177 |
|
|
{
|
| 178 |
|
|
if (! -f "$d/JOB_CREATE_LOG.txt")
|
| 179 |
|
|
{
|
| 180 |
|
|
$result{$sitename}{C} ||= 0;
|
| 181 |
|
|
$result{$sitename}{C}++;
|
| 182 |
|
|
next;
|
| 183 |
|
|
}
|
| 184 |
|
|
|
| 185 |
|
|
my $f = (<$d/crab_*/share/db/jobs>)[0];
|
| 186 |
|
|
next if ! defined $f;
|
| 187 |
|
|
|
| 188 |
|
|
foreach my $status (split(/\n/, &input($f) || ''))
|
| 189 |
|
|
{
|
| 190 |
|
|
my @statusarray = split("\;", $status);
|
| 191 |
|
|
$result{$sitename}{$statusarray[1]} ||= 0;
|
| 192 |
|
|
$result{$sitename}{$statusarray[1]}++;
|
| 193 |
|
|
}
|
| 194 |
|
|
}
|
| 195 |
|
|
}
|
| 196 |
|
|
|
| 197 |
|
|
return %result;
|
| 198 |
|
|
}
|
| 199 |
|
|
|
| 200 |
|
|
sub idle
|
| 201 |
|
|
{
|
| 202 |
|
|
my ($self, @pending) = @_;
|
| 203 |
|
|
eval {
|
| 204 |
|
|
# Get status of how busy the sites are. We obtain this only once
|
| 205 |
|
|
# in order to not favour datasets "early on" in the list.
|
| 206 |
|
|
my %sitestats = $self->getSiteStatus ();
|
| 207 |
|
|
if (keys %sitestats)
|
| 208 |
|
|
{
|
| 209 |
|
|
my @load;
|
| 210 |
|
|
foreach my $site (sort keys %sitestats)
|
| 211 |
|
|
{
|
| 212 |
|
|
push (@load, "$site" . join("", map { " $_=$sitestats{$site}{$_}" }
|
| 213 |
|
|
sort keys %{$sitestats{$site}}));
|
| 214 |
|
|
}
|
| 215 |
|
|
&logmsg ("current load: ", join ("; ", @load));
|
| 216 |
|
|
}
|
| 217 |
|
|
|
| 218 |
|
|
# Invoke links to fetch a formatted web page of published datasets.
|
| 219 |
|
|
if ( $self->{MODE} == 2 ) {
|
| 220 |
|
|
&logmsg ("DBS/DLS mode");
|
| 221 |
|
|
my $cmd = "python2.2 /data/CrabV1/COMP/JOBROBOT/DBSlistDataset.py";
|
| 222 |
|
|
my $out = `$cmd`;
|
| 223 |
|
|
foreach my $line (split(/\n/, $out))
|
| 224 |
|
|
{
|
| 225 |
|
|
&timeStart($self->{STARTTIME});
|
| 226 |
|
|
# Find out what was published and what we would like to do with it
|
| 227 |
|
|
my ($owner, $dataset, $site, $events) = split(/\//, $line);
|
| 228 |
|
|
$self->createTask($dataset, $owner, $events, $site, 2, %sitestats);
|
| 229 |
|
|
}
|
| 230 |
|
|
} else {
|
| 231 |
|
|
&logmsg ("RefDB/PubDB mode");
|
| 232 |
|
|
my $cmd = "links @{$self->{LINKS_OPTS}} -dump '$self->{URL}'";
|
| 233 |
|
|
open (PUBLISHED, "$cmd 2>/dev/null |") or die "cannot run `$cmd': $!\n";
|
| 234 |
|
|
while (<PUBLISHED>)
|
| 235 |
|
|
{
|
| 236 |
|
|
&timeStart($self->{STARTTIME});
|
| 237 |
|
|
chomp; next if ! /_/; s/\|/ /g; s/^\s+//; s/\s+$//;
|
| 238 |
|
|
# Find out what was published and what we would like to do with it
|
| 239 |
|
|
my ($dataset, $owner, $events, $site, $proto) = split(/\s+/, $_);
|
| 240 |
|
|
$self->createTask($dataset, $owner, $events, $site, 1, %sitestats);
|
| 241 |
|
|
}
|
| 242 |
|
|
close (PUBLISHED);
|
| 243 |
|
|
}
|
| 244 |
|
|
};
|
| 245 |
|
|
do { chomp ($@); &alert ($@); } if $@;
|
| 246 |
|
|
|
| 247 |
|
|
$self->nap ($self->{WAITTIME});
|
| 248 |
|
|
}
|
| 249 |
|
|
|
| 250 |
|
|
sub createTask()
|
| 251 |
|
|
{
|
| 252 |
|
|
|
| 253 |
|
|
my ($self, $dataset, $owner, $events, $site, $mode, %sitestats) = @_;
|
| 254 |
|
|
|
| 255 |
|
|
# Marco
|
| 256 |
|
|
next if ($self->{DATASET} && $dataset !~ /$self->{DATASET}/);
|
| 257 |
|
|
next if ($self->{OWNER} && $owner !~ /$self->{OWNER}/);
|
| 258 |
|
|
# Marco
|
| 259 |
|
|
my ($app, $tiers, $rc, $nevents, $output);
|
| 260 |
|
|
my $whitelist = $self->{WHITELIST}->{$site} || '.';
|
| 261 |
|
|
next if ($self->{IGNORE_REGEXP} && $site =~ /$self->{IGNORE_REGEXP}/);
|
| 262 |
|
|
next if ($self->{ACCEPT_REGEXP} && $site !~ /$self->{ACCEPT_REGEXP}/);
|
| 263 |
|
|
|
| 264 |
|
|
if ($dataset =~ /MBforPU/) {
|
| 265 |
|
|
next;
|
| 266 |
|
|
} elsif ($owner =~ /Hit/) {
|
| 267 |
|
|
$rc = "orcarc.read.simhits";
|
| 268 |
|
|
$app = "ExSimHitStatistics";
|
| 269 |
|
|
$tiers = "Hit";
|
| 270 |
|
|
$output = "simhits.aida";
|
| 271 |
|
|
} elsif ($owner =~ /DST/) {
|
| 272 |
|
|
next;
|
| 273 |
|
|
if (rand(1) > 1.75) {
|
| 274 |
|
|
$rc = "orcarc.root.dst";
|
| 275 |
|
|
$app = "ExRootAnalysisDST";
|
| 276 |
|
|
$output = "test.root";
|
| 277 |
|
|
} else {
|
| 278 |
|
|
$rc = "orcarc.read.dst";
|
| 279 |
|
|
$app = "ExDSTStatistics";
|
| 280 |
|
|
$output = "dststatistics.aida";
|
| 281 |
|
|
}
|
| 282 |
|
|
$tiers = "DST,Digi,Hit";
|
| 283 |
|
|
} else {
|
| 284 |
|
|
if (rand(1) > 1.75) {
|
| 285 |
|
|
$rc = "orcarc.root.digis";
|
| 286 |
|
|
$app = "ExRootAnalysisDigi";
|
| 287 |
|
|
$output = "test.root";
|
| 288 |
|
|
} else {
|
| 289 |
|
|
$rc = "orcarc.read.digis";
|
| 290 |
|
|
$app = "ExDigiStatistics";
|
| 291 |
|
|
$output = "digistatistics.aida";
|
| 292 |
|
|
}
|
| 293 |
|
|
$tiers = "Digi,Hit";
|
| 294 |
|
|
}
|
| 295 |
|
|
|
| 296 |
|
|
# Find out what is already pending for this task. First find all
|
| 297 |
|
|
# existing tasks in the repository, the latest generation.
|
| 298 |
|
|
my $datestamp = strftime ("%y%m%d", gmtime(time()));
|
| 299 |
|
|
my $taskdir = "$self->{TASKREPO}/$datestamp/$site/$app";
|
| 300 |
|
|
# OLI: shorten path for condor_g (restriction to 256 characters)
|
| 301 |
|
|
# my $taskbase = "$datestamp.$site.$app.$dataset.$owner";
|
| 302 |
|
|
my $taskbase = "$dataset.$owner";
|
| 303 |
|
|
my @existing = sort { $a <=> $b } map { /.*\.(\d+)$/ } <$taskdir/$taskbase.*>;
|
| 304 |
|
|
my $curgen = pop(@existing) || 0;
|
| 305 |
|
|
my $nextgen = $curgen + 1;
|
| 306 |
|
|
|
| 307 |
|
|
# If the site isn't too busy already, ignore.
|
| 308 |
|
|
my $pending = ($sitestats{$site}{S} || 0);
|
| 309 |
|
|
$pending += ($sitestats{$site}{C} || 0);
|
| 310 |
|
|
next if $pending > $self->{MAX_SITE_QUEUE};
|
| 311 |
|
|
|
| 312 |
|
|
# OK to create the task if enough time has passed from previous
|
| 313 |
|
|
# task creation, or there is no previous task.
|
| 314 |
|
|
if (! -f "$taskdir/$taskbase.$curgen/crab.cfg"
|
| 315 |
|
|
|| (((stat("$taskdir/$taskbase.$curgen/crab.cfg"))[9]
|
| 316 |
|
|
< time() - $events * $self->{SECS_PER_EVENT})))
|
| 317 |
|
|
{
|
| 318 |
|
|
my $mydir = $0; $mydir =~ s|/[^/]+$||;
|
| 319 |
|
|
my $drop = sprintf("%s.%03d", $taskbase, $nextgen);
|
| 320 |
|
|
my $ret = &runcmd ("$mydir/CrabJobs", "-app", $app,
|
| 321 |
|
|
"-jobevents", $self->{NEVENT},
|
| 322 |
|
|
"-orcarc", "$mydir/$rc",
|
| 323 |
|
|
"-owner", $owner,
|
| 324 |
|
|
"-dataset", $dataset,
|
| 325 |
|
|
"-tiers", $tiers,
|
| 326 |
|
|
"-whitelist", $whitelist,
|
| 327 |
|
|
"-name", "$taskdir/$drop",
|
| 328 |
|
|
"-output", $output,
|
| 329 |
|
|
"-jobtype", $self->{JOBTYPE},
|
| 330 |
|
|
"-scheduler", $self->{SCHEDULER},
|
| 331 |
|
|
"-mode", $mode);
|
| 332 |
|
|
die "$drop: failed to create task: @{[&runerror($ret)]}\n" if $ret;
|
| 333 |
|
|
|
| 334 |
|
|
&output ("$taskdir/$drop/TASK_INIT.txt",
|
| 335 |
|
|
&mytimeofday () . "\n");
|
| 336 |
|
|
|
| 337 |
|
|
my $dropdir = "$self->{WORKDIR}/$drop";
|
| 338 |
|
|
mkdir "$dropdir" || die "$dropdir: cannot create: $!\n";
|
| 339 |
|
|
if (&output ("$dropdir/task", "$taskdir/$drop"))
|
| 340 |
|
|
{
|
| 341 |
|
|
&touch ("$dropdir/done");
|
| 342 |
|
|
$self->relayDrop ($drop);
|
| 343 |
|
|
&logmsg("stats: $drop @{[&formatElapsedTime($self->{STARTTIME})]} success");
|
| 344 |
|
|
}
|
| 345 |
|
|
else
|
| 346 |
|
|
{
|
| 347 |
|
|
&alert ("$drop: failed to create drop");
|
| 348 |
|
|
&rmtree ([ "$self->{WORKDIR}/$drop" ]);
|
| 349 |
|
|
}
|
| 350 |
|
|
}
|
| 351 |
|
|
}
|
