rbbt-util 5.28.12 → 5.29.4

data/lib/rbbt/hpc/orchestrate.rb

@@ -0,0 +1,111 @@
+require 'rbbt/workflow/util/orchestrator'
+module HPC
+  module SLURM
+
+    def self.job_rules(rules, job)
+      workflow = job.workflow.to_s
+      task_name = job.task_name.to_s
+      defaults = rules["defaults"] || {}
+
+      job_rules = IndiferentHash.setup(defaults.dup)
+
+      rules["chains"].each do |name,info|
+        IndiferentHash.setup(info)
+        chain_tasks = info[:tasks].split(/,\s*/)
+
+        chain_tasks.each do |task|
+          task_workflow, chain_task = task.split("#")
+          chain_task, task_workflow = task_workflow, info[:workflow] if chain_task.nil? or chain_tasks.empty?
+          job_rules["chain_tasks"] ||= {}
+          job_rules["chain_tasks"][task_workflow] ||= []
+          job_rules["chain_tasks"][task_workflow]  << chain_task
+          next unless task_name == chain_task.to_s && workflow == task_workflow.to_s
+          config_keys = job_rules.delete :config_keys
+          job_rules = IndiferentHash.setup(job_rules.merge(info)) 
+          if config_keys
+            config_keys.gsub!(/,\s+/,',') 
+            job_rules[:config_keys] = job_rules[:config_keys] ? config_keys + "," + job_rules[:config_keys] : config_keys
+          end
+        end
+
+        if job_rules["chain_tasks"][workflow] && job_rules["chain_tasks"][workflow].include?(task_name)
+          break
+        else
+          job_rules.delete "chain_tasks" 
+        end
+      end if rules["chains"]
+
+      config_keys = job_rules.delete :config_keys
+      job_rules = IndiferentHash.setup(job_rules.merge(rules[workflow][task_name])) if rules[workflow] && rules[workflow][task_name]
+
+      if config_keys
+        config_keys.gsub!(/,\s+/,',') 
+        job_rules[:config_keys] = job_rules[:config_keys] ? config_keys + "," + job_rules[:config_keys] : config_keys
+      end
+
+      if rules["skip"] && rules["skip"][workflow]
+        job_rules["skip"] = true if rules["skip"][workflow].split(/,\s*/).include? task_name
+      end
+
+      job_rules
+    end
+
+    def self.get_job_dependencies(job, job_rules)
+      deps = job.dependencies || []
+      deps += job.input_dependencies || []
+      deps
+    end
+
+    def self.orchestrate_job(job, options, skip = false, seen = {})
+      return if job.done?
+      return unless job.path.split("/")[-4] == "jobs"
+      seen[:orchestration_target_job] ||= job
+
+      options.delete "recursive_clean"
+      options.delete "clean_task"
+      options.delete "clean"
+      options.delete "tail"
+      options.delete "printfile"
+      options.delete "detach"
+
+      rules = YAML.load(Open.read(options[:orchestration_rules])) if options[:orchestration_rules]
+      rules ||= {}
+      IndiferentHash.setup(rules)
+
+      job_rules = self.job_rules(rules, job)
+
+      deps = get_job_dependencies(job, job_rules)
+
+      dep_ids = deps.collect do |dep|
+        skip_dep = job_rules["chain_tasks"] &&
+          job_rules["chain_tasks"][job.workflow.to_s] && job_rules["chain_tasks"][job.workflow.to_s].include?(job.task_name.to_s)  &&
+          job_rules["chain_tasks"][dep.workflow.to_s] && job_rules["chain_tasks"][dep.workflow.to_s].include?(dep.task_name.to_s) 
+
+        deps = seen[dep.path] ||= self.orchestrate_job(dep, options, skip_dep, seen)
+        if job.canfail_paths.include? dep.path
+          [deps].flatten.collect{|id| ['canfail', id] * ":"}
+        else
+          deps
+        end
+      end.flatten.compact.uniq
+
+      skip = true if job_rules[:skip]
+      return dep_ids if skip and seen[:orchestration_target_job] != job
+
+      job_rules.delete :chain_tasks
+      job_rules.delete :tasks
+      job_rules.delete :workflow
+      
+      config_keys = job_rules.delete(:config_keys)
+
+      job_options = IndiferentHash.setup(options.merge(job_rules).merge(:slurm_dependencies => dep_ids))
+      job_options.delete :orchestration_rules
+      if config_keys
+        config_keys.gsub!(/,\s+/,',') 
+        job_options[:config_keys] = job_options[:config_keys] ? config_keys + "," + job_options[:config_keys] : config_keys
+      end
+
+      run_job(job, job_options)
+    end
+  end
+end

data/lib/rbbt/hpc/slurm.rb

@@ -0,0 +1,592 @@
+module HPC
+  class SBATCH < Exception; 
+    attr_accessor :directory
+    def initialize(directory)
+      @directory = directory
+    end
+  end
+
+  module SLURM
+
+    def self.template(args, options = {})
+
+      development      = options.delete :drbbt
+      singularity      = options.delete :singularity
+      contain          = options.delete :contain
+      sync             = options.delete :sync
+      user_group       = options.delete :user_group
+      contain_and_sync = options.delete :contain_and_sync
+      wipe_container   = options.delete :wipe_container
+      copy_image       = options.delete :copy_image
+      exclusive        = options.delete :exclusive
+      highmem          = options.delete :highmem
+
+      queue            = options.delete(:queue) || 'bsc_ls'
+      task_cpus        = options.delete(:task_cpus) || 1
+      nodes            = options.delete(:nodes) || 1
+      time             = options.delete(:time) || "0:00:10"
+
+      inputs_dir       = options.delete :inputs_dir
+      config_keys      = options.delete :config_keys
+
+      user = ENV['USER'] || `whoami`.strip
+      group = File.basename(File.dirname(ENV['HOME']))
+
+      if contain_and_sync
+        contain = "/scratch/tmp/rbbt-#{user}" if contain.nil?
+        sync = "~/.rbbt/var/jobs" if sync.nil?
+        wipe_container = "post" if wipe_container.nil?
+      end
+
+      contain = nil if contain == "" || contain == "none"
+      sync = nil if sync == "" || sync == "none"
+
+      contain = File.expand_path(contain) if contain
+
+      name = options[:name] ||= Misc.obj2digest({:options => options.collect{|k,v| [k,v]}.sort_by{|k,v| k.to_s }, :args => args})
+      options.delete(:name)
+      slurm_basedir = options[:slurm_basedir] ||= File.expand_path(File.join('~/rbbt-slurm', name)) if slurm_basedir.nil?
+      options.delete(:slurm_basedir)
+
+      rbbt_cmd = args.reject{|e| e == '--' }.collect{|e| e.include?(" ")? '"' + e + '"' : e } * " "
+
+      rbbt_cmd += " "  << options.collect do |o,v|
+        o = o.to_s
+        case v
+        when TrueClass 
+          '--' << o
+        when FalseClass 
+          '--' << o << "=false"
+        else
+          ['--' << o, "'#{v.to_s.gsub("'", '\'')}'"] * " "
+        end
+      end * " "
+
+      rbbt_cmd << " --config_keys='#{config_keys.gsub("'", '\'')}'" if config_keys and not config_keys.empty?
+
+      time = Misc.format_seconds Misc.timespan(time) unless time.include? ":"
+
+
+      #{{{ PREPARE LOCAL LOGFILES
+
+      Open.mkdir slurm_basedir
+
+      fout = File.join(slurm_basedir, 'std.out')
+      ferr = File.join(slurm_basedir, 'std.err')
+      fjob = File.join(slurm_basedir, 'job.id')
+      fexit = File.join(slurm_basedir, 'exit.status')
+      fsync = File.join(slurm_basedir, 'sync.log')
+      fsyncexit = File.join(slurm_basedir, 'sync.status')
+      fcmd = File.join(slurm_basedir, 'command.slurm')
+
+      #{{{ GENERATE TEMPLATE
+
+      # HEADER
+      header =<<-EOF
+#!/bin/bash
+#SBATCH --qos="#{queue}"
+#SBATCH --job-name="#{name}"
+#SBATCH --workdir="#{Dir.pwd}"
+#SBATCH --output="#{fout}"
+#SBATCH --error="#{ferr}"
+#SBATCH --cpus-per-task="#{task_cpus}"
+#SBATCH --time="#{time}"
+#SBATCH --nodes="#{nodes}"
+      EOF
+
+      prep  = ""
+
+      if highmem
+        header +=<<-EOF
+#SBATCH --constraint=highmem 
+        EOF
+      end
+
+      if exclusive
+        header +=<<-EOF
+#SBATCH --exclusive 
+        EOF
+      end
+
+      # ENV
+      env = ""
+      env +=<<-EOF
+# Prepare env
+[[ -f ~/config/load.sh ]] && source ~/config/load.sh
+module load java
+
+# Calculate max available memory
+let "MAX_MEMORY=$SLURM_MEM_PER_CPU * $SLURM_CPUS_PER_TASK" || let MAX_MEMORY="$(grep MemTotal /proc/meminfo|grep -o "[[:digit:]]*") / 1024"
+      EOF
+
+
+      # RUN
+      run = ""
+      exec_cmd = %(env _JAVA_OPTIONS="-Xms1g -Xmx${MAX_MEMORY}m")
+
+
+      if singularity
+        #{{{ SINGULARITY
+        
+        singularity_exec = %(singularity exec -e -B $SINGULARITY_OPT_DIR:/singularity_opt/ -B /apps/)
+        
+        env +=<<-EOF
+module load intel/2018.1
+module load singularity
+PROJECTS_ROOT="/gpfs/projects/bsc26/"
+SINGULARITY_IMG="$PROJECTS_ROOT/rbbt.singularity.img"
+SINGULARITY_OPT_DIR="$PROJECTS_ROOT/singularity_opt/"
+SINGULARITY_RUBY_INLINE="$HOME/.singularity_ruby_inline"
+mkdir -p "$SINGULARITY_RUBY_INLINE"
+        EOF
+
+        if contain
+          scratch_group_dir = File.join('/gpfs/scratch/', group)
+          projects_group_dir = File.join('/gpfs/projects/', group)
+
+          prep +=<<-EOF
+
+# Prepare container dir
+CONTAINER_DIR="#{contain}"
+mkdir -p $CONTAINER_DIR/.rbbt/etc/
+
+for dir in .ruby_inline git home; do
+    mkdir -p $CONTAINER_DIR/$dir
+done
+
+for tmpd in persist_locks  produce_locks  R_sockets  sensiblewrite  sensiblewrite_locks  step_info_locks  tsv_open_locks; do
+    mkdir -p $CONTAINER_DIR/.rbbt/tmp/$tmpd
+done
+
+# Copy environment 
+cp ~/.rbbt/etc/environment $CONTAINER_DIR/.rbbt/etc/
+
+# Set search_paths
+echo "singularity: /singularity_opt/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" > $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "rbbt_user: /home/rbbt/.rbbt/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "outside_home: $CONTAINER_DIR/home/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "group_projects: #{projects_group_dir}/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "group_scratch: #{scratch_group_dir}/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "user_projects: #{projects_group_dir}/#{user}/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+echo "user_scratch: #{scratch_group_dir}/#{user}/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+          EOF
+
+          if user_group && group != user_group
+            prep +=<<-EOF
+
+# Add user_group search_path
+echo "#{user_group}: /gpfs/projects/#{user_group}/{PKGDIR}/{TOPLEVEL}/{SUBPATH}" >> $CONTAINER_DIR/.rbbt/etc/search_paths
+            EOF
+          end
+
+          if inputs_dir
+            prep +=<<-EOF
+
+# Copy inputs
+[[ -d '#{inputs_dir}' ]] && cp -R '#{inputs_dir}' $CONTAINER_DIR/inputs
+            EOF
+            rbbt_cmd = rbbt_cmd.sub(inputs_dir, "#{contain}/inputs")
+          end
+
+          if copy_image
+            prep +=<<EOF
+
+# Copy image
+rsync -avz "$SINGULARITY_IMG" "$CONTAINER_DIR/rbbt.singularity.img" 1>&2
+SINGULARITY_IMG="$CONTAINER_DIR/rbbt.singularity.img"
+EOF
+          end
+
+          if  wipe_container == "pre" || wipe_container == "both"
+            if singularity
+              prep +=<<-EOF
+
+# Clean container pre
+singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rm -Rfv .rbbt/var/jobs &>> #{fsync}
+singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rbbt system clean -f &>> #{fsync}
+singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rm -Rfv tmp/ &>> #{fsync}
+EOF
+            else
+              prep = ""
+            end
+          end
+        end
+
+        if contain
+          singularity_exec << %( -C -H "$CONTAINER_DIR" \
+-B /scratch/tmp \
+#{ group != user_group ? "-B /gpfs/projects/#{user_group}" : "" } \
+-B #{scratch_group_dir} \
+-B #{projects_group_dir} \
+-B "$SINGULARITY_RUBY_INLINE":"$CONTAINER_DIR/.ruby_inline":rw  \
+-B ~/git:"$CONTAINER_DIR/git":ro \
+#{Open.exists?('~/.rbbt/software/opt/')? '-B ~/.rbbt/software/opt/:"/opt/":ro' : '' } \
+-B ~/.rbbt:"$CONTAINER_DIR/home/":ro \
+"$SINGULARITY_IMG")
+          exec_cmd << ' TMPDIR="$CONTAINER_DIR/.rbbt/tmp" '
+        else
+          singularity_exec += %( -B "$SINGULARITY_RUBY_INLINE":"$HOME/.ruby_inline":rw "$SINGULARITY_IMG" )
+        end
+
+        if development
+          exec_cmd += " rbbt --dev='#{development}'"
+        else
+          exec_cmd += ' rbbt'
+        end
+
+        exec_cmd = singularity_exec + " " + exec_cmd
+      else
+        if development
+          exec_cmd << " " << %(~/git/rbbt-util/bin/rbbt --dev=#{development})
+        else
+          exec_cmd << " " << 'rbbt'
+        end
+
+        if contain
+          rbbt_cmd << " " << %(--workdir_all='#{contain.gsub("'", '\'')}/.rbbt/var/jobs')
+        end
+      end
+
+
+      cmd =<<-EOF
+#{exec_cmd} \\
+#{rbbt_cmd}
+EOF
+
+      header +=<<-EOF
+#CMD: #{rbbt_cmd}
+      EOF
+
+      run +=<<-EOF
+
+# Run command
+#{cmd}
+
+# Save exit status
+exit_status=$?
+
+EOF
+
+      # CODA
+      coda = ""
+      if sync
+        if singularity
+          coda +=<<-EOF
+singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rbbt system clean all -q &>> #{fsync}
+EOF
+#        else
+#          coda +=<<-EOF
+#rbbt system clean all -q &>> #{fsync}
+#EOF
+        end
+
+        if sync.include?("=>")
+          source, _sep, sync = sync.partition("=>")
+          source = source.strip
+          sync = sync.strip
+          source = File.join(File.expand_path(contain), source)
+        else
+          source = File.join(File.expand_path(contain), '.rbbt/var/jobs')
+        end
+
+        target = File.expand_path(sync)
+        coda +=<<-EOF
+
+# Sync data to target location
+mkdir -p "$(dirname '#{target}')"
+rsync -avztAXHP --copy-unsafe-links "#{source}/" "#{target}/" &>> #{fsync} 
+sync_es="$?" 
+echo $sync_es > #{fsyncexit}
+find '#{target}' -type l -ls | awk '$13 ~ /^#{target.gsub('/','\/')}/ { sub("#{source}", "#{target}", $13); print $11, $13 }' | while read A B; do rm $A; ln -s $B $A; done
+EOF
+
+        if  contain && (wipe_container == "post" || wipe_container == "both")
+          prep =<<-EOF + prep
+if ls -A '#{contain}' &> /dev/null ; then
+    echo "ERROR: Container directory not empty, refusing to wipe. #{contain}" &>> #{fsync}
+fi
+          EOF
+          if singularity
+            coda +=<<-EOF
+singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rm -v /dev/shm/sem.*.{in,out,process} /dev/shm/sem.Session-PID.*.sem 2> /dev/null >> #{fsync}
+
+
+# Clean container directory
+#if [ $exit_status == '0' -a $sync_es == '0' ]; then 
+    singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rbbt system clean -f &>> #{fsync}
+    singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rm -Rfv .rbbt/var/jobs &>> #{fsync}
+    singularity exec -e -C -H "$CONTAINER_DIR" "$SINGULARITY_IMG" rm -Rfv tmp/ &>> #{fsync}
+#else
+#    echo "ERROR: Process failed or results could not sync correctly. Contain directory not purged" &>> #{fsync}
+#fi
+EOF
+          else
+            coda +=<<-EOF
+##{exec_cmd} system clean
+if [ $exit_status == '0' -a $sync_es == '0' ]; then 
+    rm -Rfv #{contain} &>> #{fsync}
+else
+    echo "ERROR: Process failed or results could not sync correctly. Contain directory not purged" &>> #{fsync}
+fi
+EOF
+
+          end
+        end
+      end
+
+      coda +=<<-EOF
+
+# Write exit status to file
+echo $exit_status > #{fexit}
+EOF
+
+      if sync
+        coda +=<<-EOF 
+if [ "$sync_es" == '0' ]; then 
+  unset sync_es
+  exit $exit_status
+else
+  exit $sync_es
+fi
+EOF
+      else
+        coda +=<<-EOF 
+exit $exit_status
+EOF
+      end
+
+      template = [header, env, prep, run, coda] * "\n"
+
+      template
+    end
+    
+    def self.issue_template(template, options = {})
+
+      slurm_basedir = options[:slurm_basedir]
+      dependencies = options.delete :slurm_dependencies
+      dependencies = [] if dependencies.nil?
+
+      canfail_dependencies = dependencies.select{|dep| dep =~ /^canfail:(\d+)/ }.collect{|dep| dep.partition(":").last}
+      dependencies = dependencies.reject{|dep| dep =~ /^canfail:(\d+)/ }
+
+      Open.mkdir slurm_basedir
+
+      dry_run = options.delete :dry_run
+
+      fout  = File.join(slurm_basedir, 'std.out')
+      ferr  = File.join(slurm_basedir, 'std.err')
+      fjob  = File.join(slurm_basedir, 'job.id')
+      fdep  = File.join(slurm_basedir, 'dependencies.list')
+      fcfdep  = File.join(slurm_basedir, 'canfail_dependencies.list')
+      fexit = File.join(slurm_basedir, 'exit.status')
+      fsync = File.join(slurm_basedir, 'sync.log')
+      fcmd  = File.join(slurm_basedir, 'command.slurm')
+
+      job = nil
+      if options[:clean_job]
+        [fcmd, fjob, fout, ferr, fsync, fexit].each do |file|
+          Open.rm file if Open.exists? file
+        end
+      end
+
+      return if Open.exists?(fexit)
+
+      STDERR.puts Log.color(:magenta, "Issuing SLURM file: #{fcmd}")
+      STDERR.puts template
+
+      Open.write(fcmd, template) unless File.exists? fcmd
+      if File.exists?(fjob)
+        job = Open.read(fjob).to_i
+      else
+        if File.exists?(fout)
+          return
+        elsif dry_run
+          STDERR.puts Log.color(:magenta, "To execute run: ") + Log.color(:blue, "sbatch '#{slurm_basedir}/command.slurm'")
+          STDERR.puts Log.color(:magenta, "To monitor progress run (needs local rbbt): ") + Log.color(:blue, "rbbt mn --tail -w '#{slurm_basedir}'")
+          raise HPC::SBATCH, slurm_basedir
+        else
+          Open.rm fsync
+          Open.rm fexit
+          Open.rm fout
+          Open.rm ferr
+
+          Open.write(fdep, dependencies * "\n") if dependencies.any?
+          Open.write(fcfdep, canfail_dependencies * "\n") if canfail_dependencies.any?
+
+
+          dep_str = '--dependency='
+          normal_dep_str = dependencies.any? ? "afterok:" + dependencies * ":" : nil
+          canfail_dep_str = canfail_dependencies.any? ? "afterany:" + canfail_dependencies * ":" : nil
+
+          if normal_dep_str.nil? && canfail_dep_str.nil?
+            dep_str = ""
+          else
+            dep_str += [normal_dep_str, canfail_dep_str].compact * ","
+          end
+
+          job = CMD.cmd("sbatch #{dep_str} '#{fcmd}'").read.scan(/\d+/).first.to_i
+          Log.debug "SBATCH job id: #{job}"
+          Open.write(fjob, job.to_s)
+          job
+        end
+      end
+    end
+
+    def self.follow_job(slurm_basedir, tail = true)
+      fjob = File.join(slurm_basedir, 'job.id')
+      fout = File.join(slurm_basedir, 'std.out')
+      ferr = File.join(slurm_basedir, 'std.err')
+      fstatus = File.join(slurm_basedir, 'job.status')
+
+      job = Open.read(fjob).strip if Open.exists?(fjob)
+
+      if job
+        status_txt = CMD.cmd("squeue --job #{job}").read
+        STDERR.puts Log.color(:magenta, "Status [#{job.to_i}]:")
+        STDERR.puts status_txt
+        lines = status_txt.split("\n").length
+      end
+
+      if tail
+        Log.severity = 10
+        while ! File.exists? fout
+          if job
+            STDERR.puts
+            Log.clear_line(STDERR)
+            STDERR.write Log.color(:magenta, "Waiting for Output")
+            3.times do
+              STDERR.write Log.color(:magenta, ".")
+              sleep 1
+            end
+            status_txt = CMD.cmd("squeue --job #{job}").read
+            lines.times do
+              Log.clear_line(STDERR)
+            end
+            Log.clear_line(STDERR)
+            STDERR.puts Log.color(:magenta, "Status [#{job.to_i}]:")
+            STDERR.puts status_txt
+            lines = status_txt.split("\n").length
+          end
+        end
+        STDERR.puts
+        Log.clear_line(STDERR)
+        STDERR.puts Log.color(:magenta, "Output:")
+        begin
+          CMD.cmd("squeue --job #{job} > #{fstatus}")
+          out = CMD.cmd("tail -f '#{fout}'", :pipe => true) if File.exists?(fout) and not tail == :STDERR
+          err = CMD.cmd("tail -f '#{ferr}'", :pipe => true) if File.exists?(ferr)
+
+          terr = Misc.consume_stream(err, true, STDERR) if err
+          tout = Misc.consume_stream(out, true, STDOUT) if out
+
+          sleep 3 while CMD.cmd("squeue --job #{job}").read.include? job.to_s
+        rescue Aborted
+        ensure
+          begin
+            terr.exit if terr
+            tout.exit if tout
+            err.close if err
+            err.join if err
+          rescue Exception
+          end
+
+          begin
+            out.close if out
+            out.join if out
+          rescue Exception
+          end
+        end
+      end
+    end
+
+    def self.wait_for_job(slurm_basedir, time = 1)
+      fexit = File.join(slurm_basedir, 'exit.status')
+      fjob = File.join(slurm_basedir, 'job.id')
+      job = Open.read(fjob) if Open.exists?(fjob)
+
+
+      while ! Open.exists?(fexit)
+        sleep time
+      end
+    end
+
+    def self.run_job(job, options = {})
+      options = IndiferentHash.setup(options.dup)
+        
+      dry_run          = options.delete :dry_run
+      tail             = options.delete :tail
+      dependencies     = options.delete :slurm_dependencies
+      options[:jobname] = job.clean_name
+
+      workflow = job.workflow
+
+      task = Symbol === job.overriden ? job.overriden : job.task_name
+
+      if job.overriden
+        override_deps = job.rec_dependencies.
+          select{|dep| Symbol === dep.overriden }.
+          collect do |dep| 
+
+          name = [dep.workflow.to_s, dep.task_name] * "#"
+          [name, dep.path] * "="  
+        end * ","
+      end
+
+      remove_slurm_basedir = options.delete :remove_slurm_basedir 
+      slurm_basedir = options.delete :SLURM_basedir
+      slurm_basedir = "~/rbbt-slurm" if slurm_basedir.nil?
+      TmpFile.with_file(nil, remove_slurm_basedir, :tmpdir => slurm_basedir, :prefix => "SLURM_rbbt_job-") do |tmp_directory|
+        options[:slurm_basedir] ||= tmp_directory
+        slurm_basedir = options[:slurm_basedir]
+        inputs_dir = File.join(tmp_directory, 'inputs_dir')
+        saved = Step.save_job_inputs(job, inputs_dir)
+
+        if saved && saved.any?
+          options[:inputs_dir] = inputs_dir
+          cmd = ['workflow', 'task', workflow.to_s, task.to_s, '-pf', '--load_inputs', inputs_dir, '--log', (options[:log] || Log.severity).to_s]
+        else
+          cmd = ['workflow', 'task', workflow.to_s, task.to_s, '-pf', '--log', (options[:log] || Log.severity).to_s]
+        end
+
+        cmd << "--override_deps='#{override_deps.gsub("'", '\'')}'" if override_deps and not override_deps.empty?
+
+        template = self.template(cmd, options)
+        jobid = self.issue_template(template, options.merge(:slurm_basedir => slurm_basedir, :dry_run => dry_run, :slurm_dependencies => dependencies))
+
+        return jobid unless tail
+
+        t_monitor = Thread.new do
+          self.follow_job(slurm_basedir, :STDERR)
+        end
+        self.wait_for_job(slurm_basedir)
+        t_monitor.raise Aborted
+        return unless Open.read(File.join(slurm_basedir, 'exit.status')).strip == '0'
+        path = Open.read(File.join(slurm_basedir, 'std.out')).strip
+        if Open.exists?(path) && job.path != path
+          Log.info "Path of SLURM job #{path} is different from original job #{job.path}. Stablishing link."
+          Open.ln path, job.path
+          Open.ln path + '.info', job.path + '.info'  if Open.exists?(path + '.info')
+          Open.ln path + '.files', job.path + '.files' if Open.exists?(path + '.files')
+        end
+        jobid
+      end
+    end
+  end
+
+  def self.relay(job, options={})
+    options = Misc.add_defaults options, :target => 'mn1', :search_path => 'user'
+    done_deps = job.dependencies.select do |dep|
+      dep.done? 
+    end
+
+    error_deps = job.dependencies.select do |dep|
+      dep.error? && ! dep.recoverable_error?
+    end
+
+    (done_deps + error_deps).each do |dep|
+      Step.migrate(dep.path, options[:search_path], options)
+    end
+
+  end
+end
+