[padb-devel] [padb] r124 committed - Tidy up the slurm interface code, query host count for the job...

[codesite-noreply at google.com]

Revision: 124
Author: apittman
Date: Wed Aug 19 13:15:00 2009
Log: Tidy up the slurm interface code, query host count for the job
rather than the job step which is probably correct as shadow jobs
only take a job.  Make the code run cleanly with warnings enabled.

http://code.google.com/p/padb/source/detail?r=124

Modified:
  /branches/full-duplex/src/padb

=======================================
--- /branches/full-duplex/src/padb	Wed Aug 19 07:28:24 2009
+++ /branches/full-duplex/src/padb	Wed Aug 19 13:15:00 2009
@@ -2193,15 +2193,20 @@
      return @res;
  }

+# Query the process count for the "step" as that's how many
+# processes we are going to be looking for.
  sub slurm_job_to_ncpus {
      my $job   = shift;
-    my @steps = `squeue -s -o "%i %A" 2>/dev/null`;
+    my $s     = "$job." . $conf{"slurm-job-step"};
+    my @steps = `squeue -s $s -o "%i %A" 2>/dev/null`;
      return undef if ( $? != 0 );

-# The %A option is new so ensure we have the TASKS output before we  
believe what we see here...
+    # The %A option is new so ensure we have the TASKS output
+    # before we believe what we see here...
+    # Mind you %A is several years old now so if it's not there
+    # we probably can't do anything anyway.
      my $tasks;
      my $have_tasks = 0;
-    my $s          = "$job." . $conf{"slurm-job-step"};
      foreach my $step (@steps) {
          my ( $step, $cpus ) = split( " ", $step );
          $tasks      = $cpus if ( $step eq $s );
@@ -2211,19 +2216,35 @@
      return undef;
  }

+# Query the nodecount for the "job" as that is what we shall be running on.
+sub slurm_job_to_nodecount {
+    my $job  = shift;
+    my @jobs = `squeue -o "%i %D" 2>/dev/null`;
+    return undef if ( $? != 0 );
+
+    foreach my $step (@jobs) {
+        my ( $left, $right ) = split( " ", $step );
+        return $right if ( $left eq $job );
+    }
+    return undef;
+}
+
+# Query the node list for the "step" which isn't the same as the node list
+# for the job, care should be taken if using this function to ensure this
+# is correct.
+# This functions isn't used currently.
  sub slurm_job_to_nodelist {
      my $job   = shift;
-    my @steps = `squeue -s -o "%i %N" 2>/dev/null`;
+    my $s     = "$job." . $conf{"slurm-job-step"};
+    my @steps = `squeue -s $s -o "%i %N" 2>/dev/null`;
      return undef if ( $? != 0 );

-    my $hosts;
-    my $s = "$job." . $conf{"slurm-job-step"};
      foreach my $step (@steps) {
          my ( $left, $right ) = split( " ", $step );
-        $hosts = $right if ( $left eq $s );
+        return $right if ( $left eq $s );

      }
-    return $hosts;
+    return undef;
  }

  sub slurm_job_is_running {
@@ -2234,10 +2255,9 @@
  }

  sub slurm_setup_pcmd {
-    my $job   = shift;
-    my $cpus  = slurm_job_to_ncpus($job);
-    my @nodes = slurm_job_to_nodelist($job);
-    my $nc    = $#nodes + 1;
+    my $job  = shift;
+    my $cpus = slurm_job_to_ncpus($job);
+    my $nc   = slurm_job_to_nodecount($job);
      return ( "srun --jobid=$job", $cpus, $nc );
  }

@@ -6318,7 +6338,7 @@
      foreach my $proc (@procs) {
          my ( $pid, $job, $step, $local, $global ) = split( " ", $proc );
          next if ( $global eq "-" );
-        next unless ( $job == $jobid );
+        next unless ( $job eq $jobid );
          next unless ( $step == $confInner{"slurm-job-step"} );
          $gids{$pid} = $global;
      }
@@ -6380,16 +6400,19 @@

          if (%gids) {
              $vp = vp_from_pid( \%gids, \%resmgr, $pid );
-            debug $vp, "Found $vp from $pid using scontrol listpids";
+
+            # debug $vp, "Found $vp from $pid using scontrol listpids";
          }

          if ( not defined $vp ) {
              my %env = get_remote_env($pid);

-            debug undef,
+            if (0) {
+                debug undef,
  "Checking slurm pid: $pid, job $env{SLURM_JOBID}, step $env{SLURM_STEPID},  
proc $env{SLURM_PROCID}, script $script";
-            debug undef,
+                debug undef,
  "Checking  rms  pid: $pid, job $env{RMS_JOBID}, proc $env{RMS_PROCID},  
script $script";
+            }

              if ( $env{SLURM_JOBID} eq $jobid ) {
                  if ( $env{SLURM_STEPID} eq $confInner{"slurm-job-step"} ) {
@@ -6938,6 +6961,7 @@
      $confInner{"edb"}            = find_edb();
      $confInner{"minfo"}          = find_minfo();
      $confInner{"hostname"}       = hostname();
+    $confInner{"scripts"}        = "bash,sh,dash,ash,perl,xterm";

      # Local vars to help with command line parsing
      my $outerloc;