Module: HPC::LSF

Extended by:
Orchestration, TemplateGeneration
Defined in:
lib/rbbt/hpc/lsf.rb

Class Method Summary collapse

Methods included from TemplateGeneration

batch_dir_for_id, batch_options, batch_system_variables, cleanup_environment, coda, exec_cmd, execute, follow_job, header, hold_dependencies, job_queued, job_template, load_conda, load_modules, meta_data, prepare_environment, prepare_submision, rbbt_job_exec_cmd, run_job, sync_environment, wait_for_job

Methods included from Orchestration

accumulate_rules, add_batch_deps, add_config_keys, add_rules_and_consolidate, chain_batches, check_chains, get_chains, get_job_dependencies, get_recursive_job_dependencies, job_batches, job_chains, job_dependencies, job_rules, job_rules, job_workload, merge_rules, orchestrate_job, orchestration_rules, parse_chains, pb, piggyback, prepare_for_execution, task_specific_rules, workflow_rules, workload

Class Method Details

.batch_system ⇒ Object 



8
9
10
 
# File 'lib/rbbt/hpc/lsf.rb', line 8

def self.batch_system
  "LSF"
end

.batch_system_variables ⇒ Object 



12
13
14
15
16
17
18
19
20
21
22
 
# File 'lib/rbbt/hpc/lsf.rb', line 12

def self.batch_system_variables
  <<-EOF
let TOTAL_PROCESORS="$(cat /proc/cpuinfo|grep ^processor |wc -l)"
let MAX_MEMORY_DEFAULT="$(grep MemTotal /proc/meminfo|grep -o "[[:digit:]]*") / ( (1024 * $TOTAL_PROCESORS) / $LSB_MAX_NUM_PROCESSORS )"
[ ! -z $LSB_MAX_MEM_RUSAGE ] && let MAX_MEMORY="$LSB_MAX_MEM_RUSAGE" || MAX_MEMORY="$MAX_MEMORY_DEFAULT"
export MAX_MEMORY_DEFAULT
export MAX_MEMORY
export BATCH_JOB_ID=$LSF_JOBID
export BATCH_SYSTEM=#{batch_system}
  EOF
end

.header(options = {}) ⇒ Object 



24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
 
# File 'lib/rbbt/hpc/lsf.rb', line 24

def self.header(options = {})
  options        = options.dup

  queue          = Misc.process_options options, :queue
  task_cpus      = Misc.process_options options, :task_cpus
  time           = Misc.process_options options, :time
  nodes          = Misc.process_options options, :nodes
  workdir        = Misc.process_options options, :workdir
  exclusive      = Misc.process_options options, :exclusive

  batch_dir  = Misc.process_options options, :batch_dir
  batch_name = Misc.process_options options, :batch_name
  batch_name ||= File.basename(batch_dir)

  fout       = File.join(batch_dir, 'std.out')
  ferr       = File.join(batch_dir, 'std.err')

  time = Misc.format_seconds Misc.timespan(time) unless time.include? ":"

  time = time.split(":").values_at(0, 1) * ":"

  header =<<-EOF
#!/bin/bash
#BSUB -J "#{batch_name}"
#BSUB -cwd "#{workdir}"
#BSUB -oo "#{fout}"
#BSUB -eo "#{ferr}"
#BSUB -q "#{queue}"
#BSUB -n "#{task_cpus}"
#BSUB -W "#{time}"
  EOF

  header << "#BSUB -x" << "\n" if exclusive

  header
end

.job_status(job = nil) ⇒ Object 



118
119
120
121
122
123
124
 
# File 'lib/rbbt/hpc/lsf.rb', line 118

def self.job_status(job = nil)
  if job.nil?
    CMD.cmd("bjobs -w").read
  else
    CMD.cmd("bjobs -w #{job}").read
  end
end

.run_template(batch_dir, dry_run) ⇒ Object 



61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
 
# File 'lib/rbbt/hpc/lsf.rb', line 61

def self.run_template(batch_dir, dry_run)

  fout   = File.join(batch_dir, 'std.out')
  ferr   = File.join(batch_dir, 'std.err')
  fjob   = File.join(batch_dir, 'job.id')
  fdep   = File.join(batch_dir, 'dependencies.list')
  fcfdep = File.join(batch_dir, 'canfail_dependencies.list')
  fexit  = File.join(batch_dir, 'exit.status')
  fsync  = File.join(batch_dir, 'sync.log')
  fcmd   = File.join(batch_dir, 'command.batch')

  return if Open.exists?(fexit)

  STDERR.puts Log.color(:magenta, "Issuing LSF file: #{fcmd}")
  STDERR.puts Open.read(fcmd)

  if File.exist?(fjob)
    job = Open.read(fjob).to_i
  else

    dependencies = Open.read(fdep).split("\n") if File.exist? fdep
    canfail_dependencies = Open.read(fcfdep).split("\n") if File.exist? fcfdep

    normal_dep_list = dependencies && dependencies.any? ? dependencies.collect{|d| "post_done(#{d})"} : []
    canfail_dep_list = canfail_dependencies && canfail_dependencies.any? ? canfail_dependencies.collect{|d| "done(#{d})"} : []

    dep_list = normal_dep_list + canfail_dep_list

    if dep_list.any?
      dep_str = '-w "' + dep_list * " && " + '"'
    else
      dep_str = ""
    end

    cmd = "bsub #{dep_str} < '#{fcmd}'"

    if File.exist?(fout)
      return
    elsif dry_run
      STDERR.puts Log.color(:magenta, "To execute run: ") + Log.color(:blue, cmd)
      STDERR.puts Log.color(:magenta, "To monitor progress run (needs local rbbt): ") + Log.color(:blue, "rbbt lsf tail '#{batch_dir}'")
      raise HPC::BATCH_DRY_RUN, batch_dir
    else
      Open.rm fsync
      Open.rm fexit
      Open.rm fout
      Open.rm ferr


      job = CMD.cmd(cmd).read.scan(/\d+/).first.to_i
      Log.debug "BSUB job id: #{job}"
      Open.write(fjob, job.to_s)
      job
    end
  end
end