Class: WorkflowManager::FGCZDebian10CourseCluster

Inherits:
Cluster
  • Object
show all
Defined in:
lib/workflow_manager/cluster.rb

Instance Attribute Summary

Attributes inherited from Cluster

#log_dir, #name, #options

Instance Method Summary collapse

Methods inherited from Cluster

#default_node, #generate_new_job_script, #initialize, #node_list

Constructor Details

This class inherits a constructor from WorkflowManager::Cluster

Instance Method Details

#cluster_nodes ⇒ Object 



505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
 
# File 'lib/workflow_manager/cluster.rb', line 505

def cluster_nodes
  nodes = {
    'fgcz-h-900: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-900',
    'fgcz-h-901: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-901',
    'fgcz-h-902: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-902',
    'fgcz-h-903: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-903',
    'fgcz-h-904: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-904',
    'fgcz-h-905: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-905',
    'fgcz-h-906: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-906',
    'fgcz-h-907: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-907',
    'fgcz-h-908: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-908',
    'fgcz-h-909: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-909',
    'fgcz-h-910: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-910',
    'fgcz-h-911: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-911',
    'fgcz-h-912: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-912',
    'fgcz-h-913: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-913',
    'fgcz-h-914: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-914',
    'fgcz-h-915: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-915',
    'fgcz-h-916: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-916',
    'fgcz-h-917: cpu 8,mem  30 GB,scr 500G' => 'fgcz-h-917',
  }
end

#copy_commands(org_dir, dest_parent_dir, now = nil) ⇒ Object 



496
497
498
 
# File 'lib/workflow_manager/cluster.rb', line 496

def copy_commands(org_dir, dest_parent_dir, now=nil)
  commands = ["cp -r #{org_dir} #{dest_parent_dir}"]
end

#delete_command(target) ⇒ Object 



502
503
504
 
# File 'lib/workflow_manager/cluster.rb', line 502

def delete_command(target)
  command = "rm -rf #{target}"
end

#job_ends?(log_file) ⇒ Boolean

Returns:

  • (Boolean) 


471
472
473
474
475
476
477
478
479
480
481
482
 
# File 'lib/workflow_manager/cluster.rb', line 471

def job_ends?(log_file)
  log_flag = false
  IO.popen("tail -n 10 #{log_file} 2> /dev/null") do |io|
    while line=io.gets
      if line =~ /__SCRIPT END__/
        log_flag = true
        break
      end
    end
  end
  log_flag
end

#job_pending?(job_id) ⇒ Boolean

Returns:

  • (Boolean) 


483
484
485
486
487
488
489
490
491
492
493
494
495
 
# File 'lib/workflow_manager/cluster.rb', line 483

def job_pending?(job_id)
 qstat_flag = false
  IO.popen('squeue') do |io|
    while line=io.gets
      jobid, partition, name, user, state, *others = line.chomp.split
      if jobid.strip == job_id and state =~ /PD/
        qstat_flag = true
        break
      end
    end
  end
  qstat_flag
end

#job_running?(job_id) ⇒ Boolean

Returns:

  • (Boolean) 


456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
 
# File 'lib/workflow_manager/cluster.rb', line 456

def job_running?(job_id)
 qstat_flag = false
  IO.popen('squeue') do |io|
    while line=io.gets
      # ["JOBID", "PARTITION", "NAME", "USER", "ST", "TIME", "NODES", "NODELIST(REASON)"]
      # ["206", "employee", "test.sh", "masaomi", "R", "0:03", "1", "fgcz-h-030"]
      jobid, partition, name, user, state, *others = line.chomp.split
      if jobid.strip == job_id and state == 'R'
        qstat_flag = true
        break
      end
    end
  end
  qstat_flag
end

#kill_command(job_id) ⇒ Object 



499
500
501
 
# File 'lib/workflow_manager/cluster.rb', line 499

def kill_command(job_id)
  command = "scancel #{job_id}"
end

#submit_job(script_file, script_content, option = '') ⇒ Object 



439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
 
# File 'lib/workflow_manager/cluster.rb', line 439

def submit_job(script_file, script_content, option='')
  if script_name = File.basename(script_file) and script_name =~ /\.sh/
    script_name = script_name.split(/\.sh/).first + ".sh"
    new_job_script = generate_new_job_script(script_name, script_content)
    new_job_script_base = File.basename(new_job_script)
    log_file = File.join(@log_dir, new_job_script_base + "_o.log")
    err_file = File.join(@log_dir, new_job_script_base + "_e.log")
    command = "g-sub -o #{log_file} -e #{err_file} -q course #{option} #{new_job_script}"
    job_id = `#{command}`
    job_id = job_id.chomp.split.last
    [job_id, log_file, command]
  else
    err_msg = "FGCZDebian10CourseCluster#submit_job, ERROR: script_name is not *.sh: #{File.basename(script_file)}"
    warn err_msg
    raise err_msg
  end
end