Back to num_required_workers' meaning "how many extra workers we need to add...

Back to num_required_workers' meaning "how many extra workers we need to add to this analysis"; fixing a scheduling bug/oversensitivity to manual change of batch_size

modules/Bio/EnsEMBL/Hive/AnalysisStats.pm

@@ -134,9 +134,7 @@ sub num_running_workers {
     return $self->{'_num_running_workers'};
 }
 
-# NB: the meaning of this field changes from "how many workers we need to add" to "how many workers we actually need"
-
-sub num_required_workers {
+sub num_required_workers {      # NB: the meaning of this field is, again, "how many extra workers we need to add"
     my $self = shift;
     $self->{'_num_required_workers'} = shift if(@_);
     return $self->{'_num_required_workers'};

modules/Bio/EnsEMBL/Hive/DBSQL/AnalysisStatsAdaptor.pm

@@ -90,7 +90,7 @@ sub fetch_all_by_suitability_rc_id_meadow_type {
         # the ones that clearly have work to do:
         #
     my $primary_results = $self->_generic_fetch(
-        "ast.num_running_workers<ast.num_required_workers AND ast.status in ('READY', 'WORKING')" ,
+        "ast.num_required_workers>0 AND ast.status in ('READY', 'WORKING')" ,
         $join ,
         'ORDER BY a.priority DESC, ' . ( ($self->dbc->driver eq 'sqlite') ? 'RANDOM()' : 'RAND()' ),
     );

modules/Bio/EnsEMBL/Hive/Queen.pm

@@ -250,7 +250,7 @@ sub specialize_new_worker {
             if($stats->status eq 'BLOCKED') {
                 die "Analysis is BLOCKED, can't specialize a worker";
             }
-            if($stats->num_running_workers >= $stats->num_required_workers) {
+            if($stats->num_required_workers <= 0) {
                 die "Analysis doesn't require extra workers at the moment";
             }
             if($stats->status eq 'DONE') {
@@ -621,8 +621,7 @@ sub synchronize_AnalysisStats {
     my $scheduling_allowed    =  ( !defined( $analysisStats->hive_capacity ) or $analysisStats->hive_capacity )
                               && ( !defined( $analysis->analysis_capacity  ) or $analysis->analysis_capacity  );
     my $required_workers    = $scheduling_allowed
-                            && POSIX::ceil( ($analysisStats->ready_job_count() + $analysisStats->inprogress_job_count())
-                                            / $analysisStats->get_or_estimate_batch_size() );
+                            && POSIX::ceil( $analysisStats->ready_job_count() / $analysisStats->get_or_estimate_batch_size() );
     $analysisStats->num_required_workers( $required_workers );
 
 
@@ -833,7 +832,7 @@ sub suggest_analysis_to_specialize_by_rc_id_meadow_type {
 
             #synchronize and double check that it can be run:
         $self->safe_synchronize_AnalysisStats($stats);
-        return $stats if( ($stats->status ne 'BLOCKED') and ($stats->status ne 'SYNCHING') and ($stats->num_running_workers < $stats->num_required_workers) );
+        return $stats if( ($stats->status ne 'BLOCKED') and ($stats->status ne 'SYNCHING') and ($stats->num_required_workers > 0) );
     }
 
     return undef;

modules/Bio/EnsEMBL/Hive/Scheduler.pm

@@ -119,7 +119,7 @@ sub schedule_workers {
         next if($analysis_stats->status eq 'BLOCKED');
 
             # getting the initial worker requirement for this analysis (may be stale if not sync'ed recently)
-        my $extra_workers_this_analysis = $analysis_stats->num_required_workers - $analysis_stats->num_running_workers;
+        my $extra_workers_this_analysis = $analysis_stats->num_required_workers;
 
             # if this analysis doesn't require any extra workers - just skip it:
         next if ($extra_workers_this_analysis <= 0);