Class: WorkflowManager::FGCZDevian10Cluster

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 



430
431
432
433
434
435
 
# File 'lib/workflow_manager/cluster.rb', line 430

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 400G' => 'fgcz-h-901',
  }
end

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



414
415
416
417
418
419
420
421
422
423
 
# File 'lib/workflow_manager/cluster.rb', line 414

def copy_commands(org_dir, dest_parent_dir, now=nil, queue="light")
  commands = if now == "force"
               target_file = File.join(dest_parent_dir, File.basename(org_dir))
               ["g-req copynow -f #{org_dir} #{dest_parent_dir}"]
             elsif now
               ["g-req copynow #{org_dir} #{dest_parent_dir}"]
             else
               ["g-req -w copy #{org_dir} #{dest_parent_dir}"]
             end
end

#delete_command(target) ⇒ Object 



427
428
429
 
# File 'lib/workflow_manager/cluster.rb', line 427

def delete_command(target)
  command = "g-req remove #{target}"
end

#job_ends?(log_file) ⇒ Boolean

Returns:

  • (Boolean) 


389
390
391
392
393
394
395
396
397
398
399
400
 
# File 'lib/workflow_manager/cluster.rb', line 389

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) 


401
402
403
404
405
406
407
408
409
410
411
412
413
 
# File 'lib/workflow_manager/cluster.rb', line 401

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) 


374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
 
# File 'lib/workflow_manager/cluster.rb', line 374

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 



424
425
426
 
# File 'lib/workflow_manager/cluster.rb', line 424

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

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



355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
 
# File 'lib/workflow_manager/cluster.rb', line 355

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}"
    #command = "sbatch -o #{log_file} -e #{err_file} #{new_job_script}"
    job_id = `#{command}`
    #job_id = job_id.match(/Your job (\d+) \(/)[1]
    job_id = job_id.chomp.split.last
    [job_id, log_file, command]
  else
    err_msg = "FGCZDevian10Cluster#submit_job, ERROR: script_name is not *.sh: #{File.basename(script_file)}"
    warn err_msg
    raise err_msg
  end
end