add tracker

tracker/rabit_demo.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 """
 This is the demo submission script of rabit for submitting jobs in local machine
 """

tracker/rabit_hadoop_streaming.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 """
 Deprecated
 

tracker/rabit_mpi.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 """
 Submission script to submit rabit jobs using MPI
 """

tracker/rabit_sge.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 """
 Submit rabit jobs to Sun Grid Engine
 """
@@ -38,7 +38,7 @@ else:
     
 runscript = '%s/runrabit.sh' % args.logdir
 fo = open(runscript, 'w')
-fo.write('\"$@\"')
+fo.write('\"$@\"\n')
 fo.close()
 #
 # submission script using MPI
@@ -63,7 +63,7 @@ def sge_submit(nslave, worker_args, worker_envs):
     cmd += ' %s %s' % (runscript, ' '.join(args.command + worker_args))
     print cmd
     subprocess.check_call(cmd, shell = True)
-    print 'Waiting for the jobs to get up...' % args.jobname
+    print 'Waiting for the jobs to get up...'
 
 # call submit, with nslave, the commands to run each job and submit function
 tracker.submit(args.nworker, [], fun_submit = sge_submit, verbose = args.verbose)

tracker/rabit_tracker.py

@@ -13,6 +13,7 @@ import socket
 import struct
 import subprocess
 import random
+import time
 from threading import Thread
 
 """
@@ -258,6 +259,7 @@ class Tracker:
                     job_map[s.jobid] = rank
                 if len(todo_nodes) == 0:
                     self.log_print('@tracker All of %d nodes getting started' % nslave, 2)
+                    self.start_time = time.time()
             s.assign_rank(rank, wait_conn, tree_map, parent_map, ring_map)
             if s.cmd != 'start':                
                 self.log_print('Recieve %s signal from %d' % (s.cmd, s.rank), 1)
@@ -266,6 +268,8 @@ class Tracker:
             if s.wait_accept > 0:
                 wait_conn[rank] = s
         self.log_print('@tracker All nodes finishes job', 2)
+        self.end_time = time.time()
+        self.log_print('@tracker %s secs between node start and job finish' % str(self.end_time - self.start_time), 2)
 
 def submit(nslave, args, fun_submit, verbose, hostIP = 'auto'):
     master = Tracker(verbose = verbose, hostIP = hostIP)

tracker/rabit_yarn.py

@@ -1,4 +1,4 @@
-#!/usr/bin/python
+#!/usr/bin/env python
 """
 This is a script to submit rabit job via Yarn
 rabit will run as a Yarn application