Queen.pm 30.2 KB
Newer Older
Jessica Severin's avatar
Jessica Severin committed
1 2 3 4 5 6
#
# You may distribute this module under the same terms as perl itself

=pod 

=head1 NAME
7

8
  Bio::EnsEMBL::Hive::Queen
Jessica Severin's avatar
Jessica Severin committed
9 10

=head1 SYNOPSIS
11

12
  The Queen of the Hive based job control system
Jessica Severin's avatar
Jessica Severin committed
13 14

=head1 DESCRIPTION
15

16 17 18 19 20 21 22 23 24 25 26 27
  The Queen of the Hive based job control system is responsible to 'birthing' the
  correct number of workers of the right type so that they can find jobs to do.
  It will also free up jobs of Workers that died unexpectantly so that other workers
  can claim them to do.

  Hive based processing is a concept based on a more controlled version
  of an autonomous agent type system.  Each worker is not told what to do
  (like a centralized control system - like the current pipeline system)
  but rather queries a central database for jobs (give me jobs).

  Each worker is linked to an analysis_id, registers its self on creation
  into the Hive, creates a RunnableDB instance of the Analysis->module,
28 29
  gets $analysis->stats->batch_size jobs from the job table, does its work,
  creates the next layer of job entries by interfacing to
30 31 32
  the DataflowRuleAdaptor to determine the analyses it needs to pass its
  output data to and creates jobs on the next analysis database.
  It repeats this cycle until it has lived its lifetime or until there are no
33 34 35 36 37 38 39 40 41 42
  more jobs left.
  The lifetime limit is just a safety limit to prevent these from 'infecting'
  a system.

  The Queens job is to simply birth Workers of the correct analysis_id to get the
  work down.  The only other thing the Queen does is free up jobs that were
  claimed by Workers that died unexpectantly so that other workers can take
  over the work.

  The Beekeeper is in charge of interfacing between the Queen and a compute resource
43
  or 'compute farm'.  Its job is to query Queens if they need any workers and to
44 45 46
  send the requested number of workers to open machines via the runWorker.pl script.
  It is also responsible for interfacing with the Queen to identify worker which died
  unexpectantly.
Jessica Severin's avatar
Jessica Severin committed
47 48

=head1 CONTACT
49

50
  Please contact ehive-users@ebi.ac.uk mailing list with questions/suggestions.
Jessica Severin's avatar
Jessica Severin committed
51 52

=head1 APPENDIX
53

54 55
  The rest of the documentation details each of the object methods. 
  Internal methods are usually preceded with a _
Jessica Severin's avatar
Jessica Severin committed
56 57 58

=cut

59

Jessica Severin's avatar
Jessica Severin committed
60 61 62
package Bio::EnsEMBL::Hive::Queen;

use strict;
63
use POSIX;
64
use Clone 'clone';
65 66 67
use Bio::EnsEMBL::Utils::Argument;
use Bio::EnsEMBL::Utils::Exception;

68
use Bio::EnsEMBL::Hive::Utils ('destringify', 'dir_revhash');  # import 'destringify()' and 'dir_revhash()'
69
use Bio::EnsEMBL::Hive::AnalysisJob;
Jessica Severin's avatar
Jessica Severin committed
70 71
use Bio::EnsEMBL::Hive::Worker;

72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
use base ('Bio::EnsEMBL::Hive::DBSQL::ObjectAdaptor');


sub default_table_name {
    return 'worker';
}


sub default_insertion_method {
    return 'INSERT';
}


sub object_class {
    return 'Bio::EnsEMBL::Hive::Worker';
}
88

Jessica Severin's avatar
Jessica Severin committed
89

90
############################
Jessica Severin's avatar
Jessica Severin committed
91
#
92
# PUBLIC API
Jessica Severin's avatar
Jessica Severin committed
93
#
94
############################
Jessica Severin's avatar
Jessica Severin committed
95

96

Jessica Severin's avatar
Jessica Severin committed
97 98
=head2 create_new_worker

99 100 101 102
  Description: Creates an entry in the worker table,
               populates some non-storable attributes
               and returns a Worker object based on that insert.
               This guarantees that each worker registered in this Queen's hive is properly registered.
Jessica Severin's avatar
Jessica Severin committed
103
  Returntype : Bio::EnsEMBL::Hive::Worker
104
  Caller     : runWorker.pl
Jessica Severin's avatar
Jessica Severin committed
105 106 107 108

=cut

sub create_new_worker {
109
    my ($self, @args) = @_;
110

111 112 113 114 115 116
    my ($meadow_type, $meadow_name, $process_id, $exec_host, $resource_class_id, $resource_class_name,
        $no_write, $debug, $worker_log_dir, $hive_log_dir, $job_limit, $life_span, $no_cleanup, $retry_throwing_jobs, $compile_module_once) =

    rearrange([qw(meadow_type meadow_name process_id exec_host resource_class_id resource_class_name
                no_write debug worker_log_dir hive_log_dir job_limit life_span no_cleanup retry_throwing_jobs compile_module_once) ], @args);

117
    if( defined($resource_class_name) ) {
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171
        my $rc = $self->db->get_ResourceClassAdaptor->fetch_by_name($resource_class_name)
            or die "resource_class with name='$resource_class_name' could not be fetched from the database";

        $resource_class_id = $rc->dbID;
    }

    my $sql = q{INSERT INTO worker (born, last_check_in, meadow_type, meadow_name, host, process_id, resource_class_id)
              VALUES (CURRENT_TIMESTAMP, CURRENT_TIMESTAMP, ?, ?, ?, ?, ?)};

    my $sth = $self->prepare($sql);
    $sth->execute($meadow_type, $meadow_name, $exec_host, $process_id, $resource_class_id);
    my $worker_id = $self->dbc->db_handle->last_insert_id(undef, undef, 'worker', 'worker_id')
        or die "Could not create a new worker";
    $sth->finish;

    if($hive_log_dir or $worker_log_dir) {
        my $dir_revhash = dir_revhash($worker_id);
        $worker_log_dir ||= $hive_log_dir .'/'. ($dir_revhash ? "$dir_revhash/" : '') .'worker_id_'.$worker_id;

            # Note: the following die-message will not reach the log files for circular reason!
        system("mkdir -p $worker_log_dir") && die "Could not create '$worker_log_dir' because: $!";

        my $sth_add_log = $self->prepare( "UPDATE worker SET log_dir=? WHERE worker_id=?" );
        $sth_add_log->execute($worker_log_dir, $worker_id);
        $sth_add_log->finish;
    }

    my $worker = $self->fetch_by_dbID($worker_id)
        or die "Could not fetch worker with dbID=$worker_id";

    $worker->init;

    if($job_limit) {
      $worker->job_limit($job_limit);
      $worker->life_span(0);
    }

    $worker->life_span($life_span * 60)                 if($life_span);

    $worker->execute_writes(0)                          if($no_write);

    $worker->perform_cleanup(0)                         if($no_cleanup);

    $worker->debug($debug)                              if($debug);

    $worker->retry_throwing_jobs($retry_throwing_jobs)  if(defined $retry_throwing_jobs);

    $worker->compile_module_once($compile_module_once)  if(defined $compile_module_once);

    return $worker;
}


=head2 specialize_new_worker
172

173 174
  Description: If analysis_id or logic_name is specified it will try to specialize the Worker into this analysis.
               If not specified the Queen will analyze the hive and pick the most suitable analysis.
175
  Caller     : Bio::EnsEMBL::Hive::Worker
176 177 178 179 180 181 182 183

=cut

sub specialize_new_worker {
    my ($self, $worker, @args) = @_;

    my ($analysis_id, $logic_name, $job_id, $force) =
        rearrange([qw(analysis_id logic_name job_id force) ], @args);
184

185 186
    if( scalar( grep {defined($_)} ($analysis_id, $logic_name, $job_id) ) > 1) {
        die "At most one of the options {-analysis_id, -logic_name, -job_id} can be set to pre-specialize a Worker";
Leo Gordon's avatar
Leo Gordon committed
187 188
    }

189
    my ($analysis, $stats, $special_batch);
190
    my $analysis_stats_adaptor = $self->db->get_AnalysisStatsAdaptor;
Jessica Severin's avatar
Jessica Severin committed
191

192
    if($job_id or $analysis_id or $logic_name) {    # probably pre-specialized from command-line
193

194
        if($job_id) {
195 196 197
            print "resetting and fetching job for job_id '$job_id'\n";

            my $job_adaptor = $self->db->get_AnalysisJobAdaptor;
198

199 200 201 202 203 204 205 206 207
            my $job = $job_adaptor->fetch_by_dbID( $job_id )
                or die "Could not fetch job with dbID='$job_id'";
            my $job_status = $job->status();

            if($job_status =~/(CLAIMED|PRE_CLEANUP|FETCH_INPUT|RUN|WRITE_OUTPUT|POST_CLEANUP)/ ) {
                die "Job with dbID='$job_id' is already in progress, cannot run";   # FIXME: try GC first, then complain
            } elsif($job_status =~/(DONE|SEMAPHORED)/ and !$force) {
                die "Job with dbID='$job_id' is $job_status, please use -force 1 to override";
            }
208

209 210 211 212 213 214 215 216
            if(($job_status eq 'DONE') and $job->semaphored_job_id) {
                warn "Increasing the semaphore count of the dependent job";
                $job_adaptor->increase_semaphore_count_for_jobid( $job->semaphored_job_id );
            }

            my $worker_id = $worker->dbID;
            if($job = $job_adaptor->reset_or_grab_job_by_dbID($job_id, $worker_id)) {
                $special_batch = [ $job ];
217 218
                $analysis_id = $job->analysis_id;
            } else {
219
                die "Could not claim job with dbID='$job_id' for worker with dbID='$worker_id'";
220
            }
221 222 223 224 225 226 227 228 229 230 231 232 233
        }

        if($logic_name) {
            $analysis = $self->db->get_AnalysisAdaptor->fetch_by_logic_name($logic_name)
                or die "analysis with name='$logic_name' could not be fetched from the database";

            $analysis_id = $analysis->dbID;

        } elsif($analysis_id) {
            $analysis = $self->db->get_AnalysisAdaptor->fetch_by_dbID($analysis_id)
                or die "analysis with dbID='$analysis_id' could not be fetched from the database";
        }

234 235 236
        if( $worker->resource_class_id
        and $worker->resource_class_id != $analysis->resource_class_id) {
                die "resource_class of analysis ".$analysis->logic_name." is incompatible with this Worker's resource_class";
237 238
        }

239 240 241
        $stats = $analysis_stats_adaptor->fetch_by_analysis_id($analysis_id);
        $self->safe_synchronize_AnalysisStats($stats);

242
        unless($special_batch or $force) {    # do we really need to run this analysis?
243
            if($self->get_hive_current_load() >= 1.1) {
244
                $worker->cause_of_death('HIVE_OVERLOAD');
245
                die "Hive is overloaded, can't specialize a worker";
246 247
            }
            if($stats->status eq 'BLOCKED') {
248
                die "Analysis is BLOCKED, can't specialize a worker";
249 250
            }
            if($stats->num_required_workers <= 0) {
251
                die "Analysis requires 0 workers at the moment";
252 253
            }
            if($stats->status eq 'DONE') {
254
                die "Analysis is DONE, and doesn't require workers";
255 256
            }
        }
257 258
            # probably scheduled by beekeeper.pl:
    } elsif( $stats = $self->suggest_analysis_to_specialize_by_rc_id_meadow_type($worker->resource_class_id, $worker->meadow_type) ) {
259

260
        print "Queen picked analysis with dbID=".$stats->analysis_id." for the worker\n";
261 262

        $analysis_id = $stats->analysis_id;
263 264 265
    } else {
        $worker->cause_of_death('NO_ROLE');
        die "No analysis suitable for the worker was found\n";
266
    }
267

268
        # now set it in the $worker:
Jessica Severin's avatar
Jessica Severin committed
269

270
    $worker->analysis_id( $analysis_id );
Jessica Severin's avatar
Jessica Severin committed
271

272 273 274 275
    my $sth_update_analysis_id = $self->prepare( "UPDATE worker SET analysis_id=? WHERE worker_id=?" );
    $sth_update_analysis_id->execute($worker->analysis_id, $worker->dbID);
    $sth_update_analysis_id->finish;

276 277
    if($special_batch) {
        $worker->special_batch( $special_batch );
278
    } else {    # count it as autonomous worker sharing the load of that analysis:
279 280 281

        $stats->update_status('WORKING');

282
        $analysis_stats_adaptor->decrease_required_workers($worker->analysis_id);
283 284 285 286 287 288 289 290 291
    }

        # The following increment used to be done only when no specific task was given to the worker,
        # thereby excluding such "special task" workers from being counted in num_running_workers.
        #
        # However this may be tricky to emulate by triggers that know nothing about "special tasks",
        # so I am (temporarily?) simplifying the accounting algorithm.
        #
    unless( $self->db->hive_use_triggers() ) {
292
        $analysis_stats_adaptor->increase_running_workers($worker->analysis_id);
293
    }
Jessica Severin's avatar
Jessica Severin committed
294 295
}

296

Jessica Severin's avatar
Jessica Severin committed
297
sub register_worker_death {
298
    my ($self, $worker) = @_;
299

300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323
    return unless($worker);

    my $cod = $worker->cause_of_death() || 'UNKNOWN';    # make sure we do not attempt to insert a void

    my $sql = qq{UPDATE worker SET died=CURRENT_TIMESTAMP
                    ,last_check_in=CURRENT_TIMESTAMP
                    ,status='DEAD'
                    ,work_done='}. $worker->work_done . qq{'
                    ,cause_of_death='$cod'
                WHERE worker_id='}. $worker->dbID . qq{'};
    $self->dbc->do( $sql );

    if(my $analysis_id = $worker->analysis_id) {
        my $analysis_stats_adaptor = $self->db->get_AnalysisStatsAdaptor;

        unless( $self->db->hive_use_triggers() ) {
            $analysis_stats_adaptor->decrease_running_workers($worker->analysis_id);
        }

        if($cod eq 'NO_WORK') {
            $analysis_stats_adaptor->update_status($worker->analysis_id, 'ALL_CLAIMED');
        } elsif($cod eq 'UNKNOWN'
            or $cod eq 'MEMLIMIT'
            or $cod eq 'RUNLIMIT'
324
            or $cod eq 'KILLED_BY_USER'
325
            or $cod eq 'SEE_MSG'
326
            or $cod eq 'NO_ROLE'
327
            or $cod eq 'CONTAMINATED') {
328 329 330 331 332 333 334 335 336 337
                $self->db->get_AnalysisJobAdaptor->release_undone_jobs_from_worker($worker);
        }

            # re-sync the analysis_stats when a worker dies as part of dynamic sync system
        if($self->safe_synchronize_AnalysisStats($worker->analysis->stats)->status ne 'DONE') {
            # since I'm dying I should make sure there is someone to take my place after I'm gone ...
            # above synch still sees me as a 'living worker' so I need to compensate for that
            $analysis_stats_adaptor->increase_required_workers($worker->analysis_id);
        }
    }
Jessica Severin's avatar
Jessica Severin committed
338 339
}

340

341
sub check_for_dead_workers {    # scans the whole Valley for lost Workers (but ignores unreachagle ones)
342
    my ($self, $valley, $check_buried_in_haste) = @_;
Leo Gordon's avatar
Leo Gordon committed
343

344
    warn "GarbageCollector:\tChecking for lost Workers...\n";
345

346 347 348 349
    my $queen_worker_list           = $self->fetch_overdue_workers(0);
    my %mt_and_pid_to_worker_status = ();
    my %worker_status_counts        = ();
    my %mt_and_pid_to_lost_worker   = ();
Leo Gordon's avatar
Leo Gordon committed
350

351
    warn "GarbageCollector:\t[Queen:] we have ".scalar(@$queen_worker_list)." Workers alive.\n";
352

Leo Gordon's avatar
Leo Gordon committed
353 354
    foreach my $worker (@$queen_worker_list) {

355 356 357
        my $meadow_type = $worker->meadow_type;
        if(my $meadow = $valley->find_available_meadow_responsible_for_worker($worker)) {
            $mt_and_pid_to_worker_status{$meadow_type} ||= $meadow->status_of_all_our_workers;
Leo Gordon's avatar
Leo Gordon committed
358
        } else {
359
            $worker_status_counts{$meadow_type}{'UNREACHABLE'}++;
360

361
            next;   # Worker is unreachable from this Valley
Leo Gordon's avatar
Leo Gordon committed
362 363
        }

364 365 366 367 368
        my $process_id = $worker->process_id;
        if(my $status = $mt_and_pid_to_worker_status{$meadow_type}{$process_id}) { # can be RUN|PEND|xSUSP
            $worker_status_counts{$meadow_type}{$status}++;
        } else {
            $worker_status_counts{$meadow_type}{'LOST'}++;
369

370
            $mt_and_pid_to_lost_worker{$meadow_type}{$process_id} = $worker;
371
        }
372 373 374 375 376 377
    }

        # just a quick summary report:
    foreach my $meadow_type (keys %worker_status_counts) {
        warn "GarbageCollector:\t[$meadow_type Meadow:]\t".join(', ', map { "$_:$worker_status_counts{$meadow_type}{$_}" } keys %{$worker_status_counts{$meadow_type}})."\n\n";
    }
378

379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395
    while(my ($meadow_type, $pid_to_lost_worker) = each %mt_and_pid_to_lost_worker) {
        my $this_meadow = $valley->available_meadow_hash->{$meadow_type};

        if(my $lost_this_meadow = scalar(keys %$pid_to_lost_worker) ) {
            warn "GarbageCollector:\tDiscovered $lost_this_meadow lost $meadow_type Workers\n";

            my $wpid_to_cod = {};
            if($this_meadow->can('find_out_causes')) {
                $wpid_to_cod = $this_meadow->find_out_causes( keys %$pid_to_lost_worker );
                my $lost_with_known_cod = scalar(keys %$wpid_to_cod);
                warn "GarbageCollector:\tFound why $lost_with_known_cod of $meadow_type Workers died\n";
            } else {
                warn "GarbageCollector:\t$meadow_type meadow does not support post-mortem examination\n";
            }

            warn "GarbageCollector:\tReleasing the jobs\n";
            while(my ($process_id, $worker) = each %$pid_to_lost_worker) {
396
                $worker->cause_of_death( $wpid_to_cod->{$process_id} || 'UNKNOWN');
397 398
                $self->register_worker_death($worker);
            }
399 400 401
        }
    }

402
        # the following bit is completely Meadow-agnostic and only restores database integrity:
Leo Gordon's avatar
Leo Gordon committed
403
    if($check_buried_in_haste) {
404
        warn "GarbageCollector:\tChecking for Workers buried in haste...\n";
405
        my $buried_in_haste_list = $self->fetch_all_dead_workers_with_jobs();
Leo Gordon's avatar
Leo Gordon committed
406
        if(my $bih_number = scalar(@$buried_in_haste_list)) {
407
            warn "GarbageCollector:\tfound $bih_number jobs, reclaiming.\n\n";
Leo Gordon's avatar
Leo Gordon committed
408 409 410
            if($bih_number) {
                my $job_adaptor = $self->db->get_AnalysisJobAdaptor();
                foreach my $worker (@$buried_in_haste_list) {
Leo Gordon's avatar
Leo Gordon committed
411
                    $job_adaptor->release_undone_jobs_from_worker($worker);
Leo Gordon's avatar
Leo Gordon committed
412 413 414
                }
            }
        } else {
415
            warn "GarbageCollector:\tfound none\n";
Leo Gordon's avatar
Leo Gordon committed
416 417 418
        }
    }
}
Jessica Severin's avatar
Jessica Severin committed
419 420


421 422 423
    # a new version that both checks in and updates the status
sub check_in_worker {
    my ($self, $worker) = @_;
Jessica Severin's avatar
Jessica Severin committed
424

425
    $self->dbc->do("UPDATE worker SET last_check_in=CURRENT_TIMESTAMP, status='".$worker->status."', work_done='".$worker->work_done."' WHERE worker_id='".$worker->dbID."'");
Jessica Severin's avatar
Jessica Severin committed
426 427 428
}


429
=head2 reset_job_by_dbID_and_sync
430

431
  Arg [1]: int $job_id
432
  Example: 
433
    my $job = $queen->reset_job_by_dbID_and_sync($job_id);
434
  Description: 
435
    For the specified job_id it will fetch just that job, 
436 437 438
    reset it completely as if it has never run, and return it.  
    Specifying a specific job bypasses the safety checks, 
    thus multiple workers could be running the 
439
    same job simultaneously (use only for debugging).
440
  Returntype : none
441
  Exceptions :
442
  Caller     : beekeeper.pl
443 444 445

=cut

446 447
sub reset_job_by_dbID_and_sync {
    my ($self, $job_id) = @_;
448

449
    my $job_adaptor = $self->db->get_AnalysisJobAdaptor;
450
    my $job = $job_adaptor->reset_or_grab_job_by_dbID($job_id); 
451 452 453

    my $stats = $self->db->get_AnalysisStatsAdaptor->fetch_by_analysis_id($job->analysis_id);
    $self->synchronize_AnalysisStats($stats);
454 455 456
}


457 458 459 460 461 462
######################################
#
# Public API interface for beekeeper
#
######################################

463

464 465 466
    # Note: asking for Queen->fetch_overdue_workers(0) essentially means
    #       "fetch all workers known to the Queen not to be officially dead"
    #
467
sub fetch_overdue_workers {
468
    my ($self,$overdue_secs) = @_;
469

470
    $overdue_secs = 3600 unless(defined($overdue_secs));
471

472
    my $constraint = "status!='DEAD' AND ".
473
                    ( ($self->dbc->driver eq 'sqlite')
474 475 476
                        ? "(strftime('%s','now')-strftime('%s',last_check_in))>$overdue_secs"
                        : "(UNIX_TIMESTAMP()-UNIX_TIMESTAMP(last_check_in))>$overdue_secs");
    return $self->fetch_all( $constraint );
477 478
}

479 480 481

sub fetch_all_dead_workers_with_jobs {
    my $self = shift;
Leo Gordon's avatar
Leo Gordon committed
482

483
    return $self->fetch_all( "JOIN job j USING(worker_id) WHERE worker.status='DEAD' AND j.status NOT IN ('DONE', 'READY', 'FAILED', 'PASSED_ON') GROUP BY worker_id" );
Leo Gordon's avatar
Leo Gordon committed
484
}
Jessica Severin's avatar
Jessica Severin committed
485

486

487 488
=head2 synchronize_hive

Leo Gordon's avatar
Leo Gordon committed
489
  Arg [1]    : $filter_analysis (optional)
490 491
  Example    : $queen->synchronize_hive();
  Description: Runs through all analyses in the system and synchronizes
492 493
              the analysis_stats summary with the states in the job 
              and worker tables.  Then follows by checking all the blocking rules
494 495 496 497 498 499 500
              and blocks/unblocks analyses as needed.
  Exceptions : none
  Caller     : general

=cut

sub synchronize_hive {
Leo Gordon's avatar
Leo Gordon committed
501
  my $self          = shift;
Leo Gordon's avatar
Leo Gordon committed
502
  my $filter_analysis = shift; # optional parameter
503

504
  my $start_time = time();
505

Leo Gordon's avatar
Leo Gordon committed
506
  my $list_of_analyses = $filter_analysis ? [$filter_analysis] : $self->db->get_AnalysisAdaptor->fetch_all;
Leo Gordon's avatar
Leo Gordon committed
507

Leo Gordon's avatar
Leo Gordon committed
508
  print STDERR "\nSynchronizing the hive (".scalar(@$list_of_analyses)." analyses this time):\n";
Leo Gordon's avatar
Leo Gordon committed
509 510
  foreach my $analysis (@$list_of_analyses) {
    $self->synchronize_AnalysisStats($analysis->stats);
511
    print STDERR ( ($analysis->stats()->status eq 'BLOCKED') ? 'x' : 'o');
512
  }
Leo Gordon's avatar
Leo Gordon committed
513
  print STDERR "\n";
Leo Gordon's avatar
Leo Gordon committed
514

Leo Gordon's avatar
Leo Gordon committed
515
  print STDERR ''.((time() - $start_time))." seconds to synchronize_hive\n\n";
516
}
517

518

519 520 521
=head2 safe_synchronize_AnalysisStats

  Arg [1]    : Bio::EnsEMBL::Hive::AnalysisStats object
522
  Example    : $self->safe_synchronize_AnalysisStats($stats);
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539
  Description: Prewrapper around synchronize_AnalysisStats that does
               checks and grabs sync_lock before proceeding with sync.
               Used by distributed worker sync system to avoid contention.
  Exceptions : none
  Caller     : general

=cut

sub safe_synchronize_AnalysisStats {
  my $self = shift;
  my $stats = shift;

  return $stats unless($stats->analysis_id);
  return $stats if($stats->status eq 'SYNCHING');
  return $stats if($stats->status eq 'DONE');
  return $stats if($stats->sync_lock);
  return $stats if(($stats->status eq 'WORKING') and
540
                   ($stats->seconds_since_last_update < 3*60));
541 542 543 544

  # OK try to claim the sync_lock
  my $sql = "UPDATE analysis_stats SET status='SYNCHING', sync_lock=1 ".
            "WHERE sync_lock=0 and analysis_id=" . $stats->analysis_id;
545
  #print("$sql\n");
546
  my $row_count = $self->dbc->do($sql);  
547
  return $stats unless($row_count == 1);        # return the un-updated status if locked
548
  #printf("got sync_lock on analysis_stats(%d)\n", $stats->analysis_id);
549
  
550
      # since we managed to obtain the lock, let's go and perform the sync:
551
  $self->synchronize_AnalysisStats($stats);
552

553 554 555 556
  return $stats;
}


557
=head2 synchronize_AnalysisStats
558

559 560
  Arg [1]    : Bio::EnsEMBL::Hive::AnalysisStats object
  Example    : $self->synchronize($analysisStats);
561
  Description: Queries the job and worker tables to get summary counts
562 563 564 565 566
               and rebuilds the AnalysisStats object.  Then updates the
               analysis_stats table with the new summary info
  Returntype : newly synced Bio::EnsEMBL::Hive::AnalysisStats object
  Exceptions : none
  Caller     : general
567

568
=cut
569

570
sub synchronize_AnalysisStats {
571
  my $self = shift;
572
  my $analysisStats = shift;
573

574
  return $analysisStats unless($analysisStats);
575
  return $analysisStats unless($analysisStats->analysis_id);
576

577
  $analysisStats->refresh(); ## Need to get the new hive_capacity for dynamic analyses
578 579
  my $hive_capacity = $analysisStats->hive_capacity;

580 581
  if($self->db->hive_use_triggers()) {

582 583
            my $job_count = $analysisStats->ready_job_count();
            my $required_workers = $hive_capacity && POSIX::ceil( $job_count / $analysisStats->get_or_estimate_batch_size() );
584

585 586
                # adjust_stats_for_living_workers:
            if($hive_capacity > 0) {
587
                my $unfulfilled_capacity = $hive_capacity - $analysisStats->num_running_workers();
588

589 590
                if($unfulfilled_capacity < $required_workers ) {
                    $required_workers = (0 < $unfulfilled_capacity) ? $unfulfilled_capacity : 0;
591 592 593 594 595 596
                }
            }
            $analysisStats->num_required_workers( $required_workers );

  } else {
      $analysisStats->total_job_count(0);
597 598
      $analysisStats->semaphored_job_count(0);
      $analysisStats->ready_job_count(0);
599 600 601
      $analysisStats->done_job_count(0);
      $analysisStats->failed_job_count(0);
      $analysisStats->num_required_workers(0);
602

603
            # ask for analysis_id to force MySQL to use existing index on (analysis_id, status)
604
      my $sql = "SELECT analysis_id, status, count(*) FROM job WHERE analysis_id=? GROUP BY analysis_id, status";
605 606
      my $sth = $self->prepare($sql);
      $sth->execute($analysisStats->analysis_id);
607

608 609 610
      my $done_here       = 0;
      my $done_elsewhere  = 0;
      my $total_job_count = 0;
611
      while (my ($dummy_analysis_id, $status, $job_count)=$sth->fetchrow_array()) {
612
    # print STDERR "$status: $job_count\n";
613

614
        $total_job_count += $job_count;
615

616
        if($status eq 'READY') {
617
            $analysisStats->ready_job_count($job_count);
618

619
            my $required_workers = $hive_capacity && POSIX::ceil( $job_count / $analysisStats->get_or_estimate_batch_size() );
620 621 622

                # adjust_stats_for_living_workers:
            if($hive_capacity > 0) {
623
                my $unfulfilled_capacity = $hive_capacity - $self->count_running_workers( $analysisStats->analysis_id() );
624

625 626
                if($unfulfilled_capacity < $required_workers ) {
                    $required_workers = (0 < $unfulfilled_capacity) ? $unfulfilled_capacity : 0;
627
                }
628
            }
629 630
            $analysisStats->num_required_workers( $required_workers );

631 632
        } elsif($status eq 'SEMAPHORED') {
            $analysisStats->semaphored_job_count($job_count);
633
        } elsif($status eq 'DONE') {
634
            $done_here = $job_count;
635
        } elsif($status eq 'PASSED_ON') {
636 637 638
            $done_elsewhere = $job_count;
        } elsif ($status eq 'FAILED') {
            $analysisStats->failed_job_count($job_count);
639
        }
640
      } # /while
641
      $sth->finish;
642

643 644
      $analysisStats->total_job_count( $total_job_count );
      $analysisStats->done_job_count( $done_here + $done_elsewhere );
645
  } # /unless $self->{'_hive_use_triggers'}
646

647
  $analysisStats->check_blocking_control_rules();
648

649 650 651
  if($analysisStats->status ne 'BLOCKED') {
    $analysisStats->determine_status();
  }
652

653
  # $analysisStats->sync_lock(0); ## do we perhaps need it here?
654
  $analysisStats->update;  #update and release sync_lock
655

656
  return $analysisStats;
657
}
658 659


660 661 662 663 664 665 666 667 668 669 670 671 672 673
=head2 get_num_failed_analyses

  Arg [1]    : Bio::EnsEMBL::Hive::AnalysisStats object (optional)
  Example    : if( $self->get_num_failed_analyses( $my_analysis )) { do_something; }
  Example    : my $num_failed_analyses = $self->get_num_failed_analyses();
  Description: Reports all failed analyses and returns
                either the number of total failed (if no $filter_analysis was provided)
                or 1/0, depending on whether $filter_analysis failed or not.
  Returntype : int
  Exceptions : none
  Caller     : general

=cut

Leo Gordon's avatar
Leo Gordon committed
674
sub get_num_failed_analyses {
675
    my ($self, $filter_analysis) = @_;
676

677 678 679 680 681 682 683 684 685 686 687 688
    my $failed_analyses = $self->db->get_AnalysisAdaptor->fetch_all_failed_analyses();

    my $filter_analysis_failed = 0;

    foreach my $failed_analysis (@$failed_analyses) {
        print "\t##########################################################\n";
        print "\t# Too many jobs in analysis '".$failed_analysis->logic_name."' FAILED #\n";
        print "\t##########################################################\n\n";
        if($filter_analysis and ($filter_analysis->dbID == $failed_analysis)) {
            $filter_analysis_failed = 1;
        }
    }
689

690
    return $filter_analysis ? $filter_analysis_failed : scalar(@$failed_analyses);
691 692 693
}


694
sub get_hive_current_load {
695 696 697
    my $self = shift;
    my $sql = qq{
        SELECT sum(1/hive_capacity)
698 699 700 701
        FROM worker w
        JOIN analysis_stats USING(analysis_id)
        WHERE w.status!='DEAD'
        AND hive_capacity>0
702 703 704 705 706 707
    };
    my $sth = $self->prepare($sql);
    $sth->execute();
    my ($load)=$sth->fetchrow_array();
    $sth->finish;
    return ($load || 0);
708 709 710
}


711 712 713
sub count_running_workers {
    my ($self, $analysis_id) = @_;

714 715 716 717 718
    my $sql = qq{
            SELECT count(*)
            FROM worker
            WHERE status!='DEAD'
        } . ($analysis_id ? " AND analysis_id='$analysis_id'" : '');
719 720 721 722 723 724 725

    my $sth = $self->prepare($sql);
    $sth->execute();
    (my $running_workers_count)=$sth->fetchrow_array();
    $sth->finish();

    return $running_workers_count || 0;
726 727
}

728

Leo Gordon's avatar
Leo Gordon committed
729
sub get_remaining_jobs_show_hive_progress {
730
  my $self = shift;
Leo Gordon's avatar
Leo Gordon committed
731
  my $sql = "SELECT sum(done_job_count), sum(failed_job_count), sum(total_job_count), ".
732
            "sum(ready_job_count * analysis_stats.avg_msec_per_job)/1000/60/60 ".
733
            "FROM analysis_stats";
734 735
  my $sth = $self->prepare($sql);
  $sth->execute();
736
  my ($done, $failed, $total, $cpuhrs) = $sth->fetchrow_array();
737
  $sth->finish;
738 739 740 741 742 743 744

  $done   ||= 0;
  $failed ||= 0;
  $total  ||= 0;
  my $completed = $total
    ? ((100.0 * ($done+$failed))/$total)
    : 0.0;
745 746 747
  my $remaining = $total - $done - $failed;
  printf("hive %1.3f%% complete (< %1.3f CPU_hrs) (%d todo + %d done + %d failed = %d total)\n", 
          $completed, $cpuhrs, $remaining, $done, $failed, $total);
Leo Gordon's avatar
Leo Gordon committed
748
  return $remaining;
749 750
}

751

Leo Gordon's avatar
Leo Gordon committed
752
sub print_analysis_status {
Leo Gordon's avatar
Leo Gordon committed
753
    my ($self, $filter_analysis) = @_;
754

Leo Gordon's avatar
Leo Gordon committed
755
    my $list_of_analyses = $filter_analysis ? [$filter_analysis] : $self->db->get_AnalysisAdaptor->fetch_all;
Leo Gordon's avatar
Leo Gordon committed
756
    foreach my $analysis (sort {$a->dbID <=> $b->dbID} @$list_of_analyses) {
757
        $analysis->stats->print_stats();
Leo Gordon's avatar
Leo Gordon committed
758
    }
759 760 761
}


Leo Gordon's avatar
Leo Gordon committed
762
sub print_running_worker_counts {
763
    my $self = shift;
764

765 766 767 768 769 770 771
    my $sql = qq{
        SELECT logic_name, count(*)
        FROM worker w
        JOIN analysis_base USING(analysis_id)
        WHERE w.status!='DEAD'
        GROUP BY analysis_id
    };
Leo Gordon's avatar
Leo Gordon committed
772

773 774 775 776 777 778 779 780 781 782 783
    my $total_workers = 0;
    my $sth = $self->prepare($sql);
    $sth->execute();

    print "\n===== Stats of live Workers according to the Queen: ======\n";
    while((my $logic_name, my $worker_count)=$sth->fetchrow_array()) {
        printf("%30s : %d workers\n", $logic_name, $worker_count);
        $total_workers += $worker_count;
    }
    $sth->finish;
    printf("%30s : %d workers\n\n", '======= TOTAL =======', $total_workers);
784
}
785

786

787 788 789 790 791 792 793 794 795 796 797
=head2 monitor

  Arg[1]     : --none--
  Example    : $queen->monitor();
  Description: Monitors current throughput and store the result in the monitor
               table
  Exceptions : none
  Caller     : beekeepers and other external processes

=cut

Leo Gordon's avatar
Leo Gordon committed
798
sub monitor {
799 800 801 802
  my $self = shift;
  my $sql = qq{
      INSERT INTO monitor
      SELECT
803
          CURRENT_TIMESTAMP,
804
          count(*),
Leo Gordon's avatar
Leo Gordon committed
805 806 807 808 809 810
  }. ( ($self->dbc->driver eq 'sqlite')
        ? qq{ sum(work_done/(strftime('%s','now')-strftime('%s',born))),
              sum(work_done/(strftime('%s','now')-strftime('%s',born)))/count(*), }
        : qq{ sum(work_done/(UNIX_TIMESTAMP()-UNIX_TIMESTAMP(born))),
              sum(work_done/(UNIX_TIMESTAMP()-UNIX_TIMESTAMP(born)))/count(*), }
  ). qq{
811
          group_concat(DISTINCT logic_name)
812 813 814
      FROM worker w
      LEFT JOIN analysis_base USING (analysis_id)
      WHERE w.status!='DEAD'
815
  };
816 817 818 819 820
      
  my $sth = $self->prepare($sql);
  $sth->execute();
}

821

Leo Gordon's avatar
Leo Gordon committed
822 823 824 825 826 827 828 829 830 831 832 833 834 835
=head2 register_all_workers_dead

  Example    : $queen->register_all_workers_dead();
  Description: Registers all workers dead
  Exceptions : none
  Caller     : beekeepers and other external processes

=cut

sub register_all_workers_dead {
    my $self = shift;

    my $overdueWorkers = $self->fetch_overdue_workers(0);
    foreach my $worker (@{$overdueWorkers}) {
836
        $worker->cause_of_death( 'UNKNOWN' );  # well, maybe we could have investigated further...
Leo Gordon's avatar
Leo Gordon committed
837 838 839 840
        $self->register_worker_death($worker);
    }
}

841

842
sub suggest_analysis_to_specialize_by_rc_id_meadow_type {
843
    my $self                = shift;
844
    my $rc_id               = shift;
845
    my $meadow_type         = shift;
846

847
    my @suitable_analyses = @{ $self->db->get_AnalysisStatsAdaptor->fetch_all_by_suitability_rc_id_meadow_type( $rc_id, $meadow_type ) };
848

849
    foreach my $stats (@suitable_analyses) {
850

851
            #synchronize and double check that it can be run:
852
        $self->safe_synchronize_AnalysisStats($stats);
853
        return $stats if( ($stats->status ne 'BLOCKED') and ($stats->status ne 'SYNCHING') and ($stats->num_required_workers > 0) );
854
    }
855

856
    return undef;