util/qdo

   1 #! /usr/bin/env python
   2
   3 # Copyright (c) 2004-2005 The Regents of The University of Michigan
   4 # All rights reserved.
   5 #
   6 # Redistribution and use in source and binary forms, with or without
   7 # modification, are permitted provided that the following conditions are
   8 # met: redistributions of source code must retain the above copyright
   9 # notice, this list of conditions and the following disclaimer;
  10 # redistributions in binary form must reproduce the above copyright
  11 # notice, this list of conditions and the following disclaimer in the
  12 # documentation and/or other materials provided with the distribution;
  13 # neither the name of the copyright holders nor the names of its
  14 # contributors may be used to endorse or promote products derived from
  15 # this software without specific prior written permission.
  16 #
  17 # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  18 # "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  19 # LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  20 # A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  21 # OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  22 # SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  23 # LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  24 # DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  25 # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  26 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  27 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  28
  29 import sys
  30 import os
  31 import re
  32 import time
  33 import optparse
  34
  35 import pexpect
  36
  37 progname = os.path.basename(sys.argv[0])
  38
  39 usage = "%prog [options] command [command arguments]"
  40 optparser = optparse.OptionParser(usage=usage)
  41 optparser.allow_interspersed_args=False
  42 optparser.add_option('-e', dest='stderr_file',
  43                      help='command stderr output file')
  44 optparser.add_option('-o', dest='stdout_file',
  45                      help='command stdout output file')
  46 optparser.add_option('-l', dest='save_log', action='store_true',
  47                      help='save qsub output log file')
  48 optparser.add_option('-N', dest='job_name',
  49                      help='qsub job name')
  50 optparser.add_option('-q', dest='dest_queue',
  51                      help='qsub destination queue')
  52 optparser.add_option('--qwait', dest='qsub_timeout', type='int',
  53                      help='qsub queue wait timeout', default=30*60)
  54 optparser.add_option('-t', dest='cmd_timeout', type='int',
  55                      help='command execution timeout', default=600*60)
  56
  57 (options, cmd) = optparser.parse_args()
  58
  59 if cmd == []:
  60     print >>sys.stderr, "%s: missing command" % progname
  61     sys.exit(1)
  62
  63 # If we want to do this, need to add check here to make sure cmd[0] is
  64 # a valid PBS job name, else qsub will die on us.
  65 #
  66 #if not options.job_name:
  67 #    options.job_name = cmd[0]
  68
  69 cwd = os.getcwd()
  70
  71 # Deal with systems where /n is a symlink to /.automount
  72 if cwd.startswith('/.automount/'):
  73     cwd = cwd.replace('/.automount/', '/n/', 1)
  74
  75 if not cwd.startswith('/n/poolfs/'):
  76     print >>sys.stderr, "Error: current directory must be under /n/poolfs."
  77     sys.exit(1)
  78
  79 # The Shell class wraps pexpect.spawn with some handy functions that
  80 # assume the thing on the other end is a Bourne/bash shell.
  81 class Shell(pexpect.spawn):
  82     # Regexp to match the shell prompt.  We change the prompt to
  83     # something fixed and distinctive to make it easier to match
  84     # reliably.
  85     prompt_re = re.compile('qdo\$ ')
  86
  87     def __init__(self, cmd):
  88         # initialize base pexpect.spawn object
  89         try:
  90             pexpect.spawn.__init__(self, cmd)
  91         except pexpect.ExceptionPexpect, exc:
  92             print "%s:" % progname, exc
  93             sys.exit(1)
  94         # full_output accumulates the full output of the session
  95         self.full_output = ""
  96         self.quick_timeout = 15
  97         # wait for a prompt, then change it
  98         try:
  99             self.expect('\$ ', options.qsub_timeout)
 100         except pexpect.TIMEOUT:
 101             print >>sys.stderr, "%s: qsub timed out." % progname
 102             self.kill(9)
 103             self.close(wait=True)
 104             sys.exit(1)
 105         self.do_command('unset PROMPT_COMMAND; PS1="qdo$ "')
 106
 107     # version of expect that updates full_output too
 108     def expect(self, regexp, timeout = -1):
 109         pexpect.spawn.expect(self, regexp, timeout)
 110         self.full_output += self.before + self.after
 111
 112     # Just issue a command and wait for the next prompt.
 113     # Returns a string containing the output of the command.
 114     def do_bare_command(self, cmd, timeout = -1):
 115         global full_output
 116         self.sendline(cmd)
 117         # read back the echo of the command
 118         self.readline()
 119         # wait for the next prompt
 120         self.expect(self.prompt_re, timeout)
 121         output = self.before.rstrip()
 122         return output
 123
 124     # Issue a command, then query its exit status.
 125     # Returns a (string, int) tuple with the command output and the status.
 126     def do_command(self, cmd, timeout = -1):
 127         # do the command itself
 128         output = self.do_bare_command(cmd, timeout)
 129         # collect status
 130         status = int(self.do_bare_command("echo $?", self.quick_timeout))
 131         return (output, status)
 132
 133     # Check to see if the given directory exists.
 134     def dir_exists(self, dirname):
 135         (output, status) = shell.do_command('[ -d %s ]' % dirname,
 136                                             self.quick_timeout)
 137         return status == 0
 138
 139
 140 # Spawn the interactive pool job.
 141
 142 # Hack to do link on poolfs... disabled for now since
 143 # compiler/linker/library versioning problems between poolfs and
 144 # nodes.  May never work since poolfs is x86-64 and nodes are 32-bit.
 145 if False and len(cmd) > 50:
 146     shell_cmd = 'ssh -t poolfs /bin/sh -l'
 147     print "%s: running %s on poolfs" % (progname, cmd[0])
 148 else:
 149     shell_cmd = 'qsub -I -S /bin/sh'
 150     if options.job_name:
 151         shell_cmd += ' -N "%s"' % options.job_name
 152     if options.dest_queue:
 153         shell_cmd += ' -q ' + options.dest_queue
 154
 155 shell = Shell(shell_cmd)
 156
 157 try:
 158     # chdir to cwd
 159     (output, status) = shell.do_command('cd ' + cwd)
 160
 161     if status != 0:
 162         raise OSError, "Can't chdir to %s" % cwd
 163
 164     # wacky hack: sometimes scons will create an output directory then
 165     # fork a job to generate files in that directory, and the job will
 166     # get run before the directory creation propagates through NFS.
 167     # This hack looks for a '-o' option indicating an output file and
 168     # waits for the corresponding directory to appear if necessary.
 169     try:
 170         if 'cc' in cmd[0] or 'g++' in cmd[0]:
 171             output_dir = os.path.dirname(cmd[cmd.index('-o')+1])
 172         elif 'm5' in cmd[0]:
 173             output_dir = cmd[cmd.index('-d')+1]
 174         else:
 175             output_dir = None
 176     except (ValueError, IndexError):
 177         # no big deal if there's no '-o'/'-d' or if it's the final argument
 178         output_dir = None
 179
 180     if output_dir:
 181         secs_waited = 0
 182         while not shell.dir_exists(output_dir) and secs_waited < 90:
 183             time.sleep(5)
 184             secs_waited += 5
 185         if secs_waited > 30:
 186             print "waited", secs_waited, "seconds for", output_dir
 187
 188     # run command
 189     if options.stdout_file:
 190         cmd += ['>', options.stdout_file]
 191     if options.stderr_file:
 192         cmd += ['2>', options.stderr_file]
 193     try:
 194         (output, status) = shell.do_command(' '.join(cmd), options.cmd_timeout)
 195     except pexpect.TIMEOUT:
 196             print >>sys.stderr, "%s: command timed out after %d seconds." \
 197                   % (progname, options.cmd_timeout)
 198             shell.sendline('~.') # qsub/ssh termination escape sequence
 199             shell.close(wait=True)
 200             status = 3
 201     if output:
 202         print output
 203
 204 finally:
 205     # end job
 206     if shell.isalive():
 207         shell.sendline('exit')
 208         shell.expect('qsub: job .* completed\r\n')
 209         shell.close(wait=True)
 210
 211     # if there was an error, log the output even if not requested
 212     if status != 0 or options.save_log:
 213         log = file('qdo-log.' + str(os.getpid()), 'w')
 214         log.write(shell.full_output)
 215         log.close()
 216
 217 del shell
 218
 219 sys.exit(status)