def setup_tunnel(address, port):
"""
Setup tunnel to `address` and `port` assuming:
- The remote login name matches the local login name.
- `port` is available on the local host.
- 'plink' is available on Windows, 'ssh' on other platforms.
- No user interaction is required to connect via 'plink'/'ssh'.
address: string
IPv4 address to tunnel to.
port: int
Port at `address` to tunnel to.
Returns ``(local_address, local_port)``.
"""
logname = 'tunnel-%s-%d.log' % (address, port)
logname = os.path.join(os.getcwd(), logname)
stdout = open(logname, 'w')
user = getpass.getuser()
if sys.platform == 'win32': # pragma no cover
stdin = open('nul:', 'r')
args = ['plink', '-ssh', '-l', user,
'-L', '%d:localhost:%d' % (port, port), address]
else:
stdin = open('/dev/null', 'r')
args = ['ssh', '-l', user,
'-L', '%d:localhost:%d' % (port, port), address]
tunnel_proc = ShellProc(args, stdin=stdin, stdout=stdout, stderr=STDOUT)
sock = socket.socket(socket.AF_INET)
address = ('127.0.0.1', port)
for retry in range(20):
time.sleep(.5)
exitcode = tunnel_proc.poll()
if exitcode is not None:
msg = 'ssh tunnel process exited with exitcode %d,' \
' output in %s' % (exitcode, logname)
logging.error(msg)
raise RuntimeError(msg)
try:
sock.connect(address)
except socket.error as exc:
if exc.args[0] != errno.ECONNREFUSED and \
exc.args[0] != errno.ENOENT:
raise
else:
atexit.register(_cleanup_tunnel, tunnel_proc, logname)
sock.close()
return address
_cleanup_tunnel(tunnel_proc, logname)
raise RuntimeError('Timeout trying to connect through tunnel to %s'
% address)
def _get_hosts(self):
""" Return list of hostnames sorted by load. """
# Get host load information.
try:
proc = ShellProc(self._QHOST, stdout=PIPE)
except Exception as exc:
self._logger.error('%r failed: %s' % (self._QHOST, exc))
return []
lines = proc.stdout.readlines()
# Reduce to hosts we're interested in and sort by CPU-adjusted load.
loads = []
for line in lines:
if line.startswith(('HOSTNAME', '-')):
continue
hostname, arch, ncpu, load, \
memtot, memuse, swapto, swapus = line.split()
if self.pattern:
if not fnmatch.fnmatchcase(hostname, self.pattern):
continue
try:
load = float(load)
ncpu = int(ncpu)
except ValueError:
continue
loads.append((hostname, load / ncpu, ncpu))
loads = sorted(loads, key=lambda item: item[1])
# Return list of hostnames.
hosts = []
for hostname, load, ncpu in loads:
for i in range(ncpu):
hosts.append(hostname)
return hosts
def test_errormsg(self):
logging.debug('')
logging.debug('test_errormsg')
cmd = 'dir' if sys.platform == 'win32' else 'ls'
try:
proc = ShellProc(cmd, stdout='stdout', stderr='stderr')
proc.wait()
finally:
if os.path.exists('stdout'):
os.remove('stdout')
if os.path.exists('stderr'):
os.remove('stderr')
msg = proc.error_message(-signal.SIGTERM)
if sys.platform == 'win32':
self.assertEqual(msg, '')
else:
self.assertEqual(msg, ': SIGTERM')
def test_errormsg(self):
logging.debug('')
logging.debug('test_errormsg')
cmd = 'dir' if sys.platform == 'win32' else 'ls'
try:
proc = ShellProc(cmd, stdout='stdout', stderr='stderr')
proc.wait()
finally:
if os.path.exists('stdout'):
os.remove('stdout')
if os.path.exists('stderr'):
os.remove('stderr')
msg = proc.error_message(-signal.SIGTERM)
if sys.platform == 'win32':
self.assertEqual(msg, '')
else:
self.assertEqual(msg, ': SIGTERM')
def execute_command(self, command, stdin, stdout, stderr, env_vars,
poll_delay, timeout):
"""
Run `command` in a subprocess if this server's `allow_shell`
attribute is True.
command: string
Command line to be executed.
stdin, stdout, stderr: string
Filenames for the corresponding stream.
env_vars: dict
Environment variables for the command.
poll_delay: float (seconds)
Delay between polling subprocess for completion.
timeout: float (seconds)
Maximum time to wait for command completion. A value of zero
implies no timeout.
"""
self._logger.debug('execute_command %r', command)
if not self._allow_shell:
self._logger.error('attempt to execute %r by %r', command,
get_credentials().user)
raise RuntimeError('shell access is not allowed by this server')
for arg in (stdin, stdout, stderr):
if isinstance(arg, basestring):
self._check_path(arg, 'execute_command')
try:
process = ShellProc(command, stdin, stdout, stderr, env_vars)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(poll_delay, timeout)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
def _execute_local(self):
""" Run command. """
self._logger.info('executing %s...', self.command)
start_time = time.time()
self._process = \
ShellProc(self.command, self.stdin, self.stdout, self.stderr,
self.env_vars)
self._logger.debug('PID = %d', self._process.pid)
try:
return_code, error_msg = \
self._process.wait(self.poll_delay, self.timeout)
finally:
self._process.close_files()
self._process = None
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %.1f sec.', et)
return (return_code, error_msg)
def _unused_remote_port(address, port, user, identity):
""" Return a (currently) unused port on `address`, default to `port`. """
if '@' in address:
user, host = address.split('@')
else:
user = user or getpass.getuser()
host = address
if sys.platform == 'win32': # pragma no cover
cmd = ['plink', '-batch', '-ssh']
else:
cmd = ['ssh']
cmd += ['-l', user]
if identity:
cmd += ['-i', identity]
cmd += ['-x', '-T']
# FIXME: this currently won't work for Windows if ssh doesn't connect to a
# UNIX-like shell (cygwin, etc.)
code = '''"import socket
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.bind(('localhost', 0))
port = sock.getsockname()[1]
sock.close()
print 'port', port"'''
cmd += [host, 'python', '-c', code.replace('\n', ';')]
try:
proc = ShellProc(cmd,
stdout=PIPE,
stderr=PIPE,
universal_newlines=True)
except Exception as exc:
logging.warning("Can't get unused port on %s from %s (forcing %s): %s",
host, cmd, port, exc)
return port
output = proc.stdout.read()
for line in output.split('\n'):
if line.startswith('port'):
remote_port = int(line.split()[1])
logging.debug('Unused remote port %s on %s', remote_port, host)
return remote_port
else:
logging.warning(
"Can't get unused port on %s from %s (forcing %s):\n"
"[stdout]\n%s\n[stderr]\n%s", host, cmd, port, output,
proc.stderr.read())
return port
def _start_tunnel(address, port, args, user, identity, prefix):
""" Start an ssh tunnel process. """
if '@' in address:
user, host = address.split('@')
else:
user = user or getpass.getuser()
host = address
if sys.platform == 'win32': # pragma no cover
cmd = ['plink', '-batch', '-ssh']
else:
cmd = ['ssh']
cmd += ['-l', user]
if identity:
cmd += ['-i', identity]
cmd += ['-N', '-x', '-T'] # plink doesn't support '-n' (no stdin)
cmd += args + [host]
logname = '%s-%s-%s.log' % (prefix, host, port)
logname = os.path.join(os.getcwd(), logname)
stdout = open(logname, 'w')
tunnel_proc = None
try:
tunnel_proc = ShellProc(cmd, stdout=stdout, stderr=STDOUT)
except Exception as exc:
raise RuntimeError("Can't create ssh tunnel process from %s: %s"
% (cmd, exc))
time.sleep(1)
exitcode = tunnel_proc.poll()
if exitcode is not None:
raise RuntimeError('ssh tunnel process for %s:%s exited with exitcode'
' %d, output in %s'
% (address, port, exitcode, logname))
return (_cleanup_tunnel, tunnel_proc, stdout, logname, os.getpid())
def _start_tunnel(address, port, args, user, identity, prefix):
""" Start an ssh tunnel process. """
if "@" in address:
user, host = address.split("@")
else:
user = user or getpass.getuser()
host = address
if sys.platform == "win32": # pragma no cover
cmd = ["plink", "-batch", "-ssh"]
else:
cmd = ["ssh"]
cmd += ["-l", user]
if identity:
cmd += ["-i", identity]
cmd += ["-N", "-x", "-T"] # plink doesn't support '-n' (no stdin)
cmd += args + [host]
logname = "%s-%s-%s.log" % (prefix, host, port)
logname = os.path.join(os.getcwd(), logname)
stdout = open(logname, "w")
tunnel_proc = None
try:
tunnel_proc = ShellProc(cmd, stdout=stdout, stderr=STDOUT)
except Exception as exc:
raise RuntimeError("Can't create ssh tunnel process from %s: %s" % (cmd, exc))
time.sleep(1)
exitcode = tunnel_proc.poll()
if exitcode is not None:
raise RuntimeError(
"ssh tunnel process for %s:%s exited with exitcode" " %d, output in %s" % (address, port, exitcode, logname)
)
return (_cleanup_tunnel, tunnel_proc, stdout, logname, os.getpid())
def _execute_local(self):
""" Run command. """
self._logger.info("executing '%s'...", self.command)
start_time = time.time()
self._process = \
ShellProc(self.command, self.stdin, self.stdout, self.stderr,
self.env_vars)
self._logger.debug('PID = %d', self._process.pid)
try:
return_code, error_msg = \
self._process.wait(self.poll_delay, self.timeout)
finally:
self._process.close_files()
self._process = None
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %.1f sec.', et)
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-sync yes' `qsub` option is used to wait for job completion.
Other job resource keys are processed as follows:
========================= =========================
Resource Key Translation
========================= =========================
submit_as_hold -h
------------------------- -------------------------
rerunnable -r yes|no
------------------------- -------------------------
working_directory -wd `value`
------------------------- -------------------------
job_category Sets parallel environment
------------------------- -------------------------
min_cpus Sets parallel environment
------------------------- -------------------------
max_cpus Sets parallel environment
------------------------- -------------------------
min_phys_memory Ignored
------------------------- -------------------------
email -M `value`
------------------------- -------------------------
email_on_started -m b
------------------------- -------------------------
email_on_terminated -m e
------------------------- -------------------------
job_name -N `value`
------------------------- -------------------------
input_path -i `value`
------------------------- -------------------------
output_path -o `value`
------------------------- -------------------------
error_path -e `value`
------------------------- -------------------------
join_files -j yes|no
------------------------- -------------------------
reservation_id -ar `value`
------------------------- -------------------------
queue_name -q `value`
------------------------- -------------------------
priority -p `value`
------------------------- -------------------------
start_time -a `value`
------------------------- -------------------------
deadline_time Ignored
------------------------- -------------------------
accounting_id -A `value`
========================= =========================
Where `value` is the corresponding resource value.
If 'working_directory' is not specified, add ``-cwd``.
If 'input_path' is not specified, add ``-i /dev/null``.
If 'output_path' is not specified, add ``-o <remote_command>.stdout``.
If 'error_path' is not specified, add ``-j yes``.
If 'native_specification' is specified, it is added to the `qsub`
command just before 'remote_command' and 'args'.
If specified, 'job_category' is used to index into the category
map set up during allocator configuration. The mapped category
name as well as the 'min_cpus' and 'max_cpus' values are used
with the ``-pe`` qsub option.
Some resource limits are also handled:
==================== =========================
Resource Key Translation
==================== =========================
core_file_size Ignored
-------------------- -------------------------
data_seg_size Ignored
-------------------- -------------------------
file_size Ignored
-------------------- -------------------------
open_files Ignored
-------------------- -------------------------
stack_size Ignored
-------------------- -------------------------
virtual_memory Ignored
-------------------- -------------------------
cpu_time -l h_cpu= `value`
-------------------- -------------------------
wallclock_time -l h_rt= `value`
==================== =========================
Output from `qsub` itself is routed to ``qsub.out``.
"""
self.home_dir = os.path.expanduser('~')
self.work_dir = ''
cmd = list(self._QSUB)
cmd.extend(('-V', '-sync', 'yes', '-b', 'yes'))
env = None
inp, out, err = None, None, None
# Set working directory now, for possible path fixing.
try:
value = resource_desc['working_directory']
except KeyError:
pass
else:
self.work_dir = self._fix_path(value)
cmd.extend(('-wd', value))
# Process description in fixed, repeatable order.
keys = ('submit_as_hold',
'rerunnable',
'job_environment',
'email',
'email_on_started',
'email_on_terminated',
'job_name',
'input_path',
'output_path',
'error_path',
'join_files',
'reservation_id',
'queue_name',
'priority',
'start_time',
'accounting_id')
email_events = ''
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == 'submit_as_hold':
if value:
cmd.append('-h')
elif key == 'rerunnable':
cmd.extend(('-r', 'yes' if value else 'no'))
elif key == 'job_environment':
env = value
elif key == 'email':
cmd.extend(('-M', ','.join(value)))
elif key == 'email_on_started':
email_events += 'b'
elif key == 'email_on_terminated':
email_events += 'e'
elif key == 'job_name':
cmd.extend(('-N', self._jobname(value)))
elif key == 'input_path':
cmd.extend(('-i', self._fix_path(value)))
inp = value
elif key == 'output_path':
cmd.extend(('-o', self._fix_path(value)))
out = value
elif key == 'error_path':
cmd.extend(('-e', self._fix_path(value)))
err = value
elif key == 'join_files':
cmd.extend(('-j', 'yes' if value else 'no'))
if value:
err = 'yes'
elif key == 'reservation_id':
cmd.extend(('-ar', value))
elif key == 'queue_name':
cmd.extend(('-q', value))
elif key == 'priority':
cmd.extend(('-p', str(value)))
elif key == 'start_time':
cmd.extend(('-a', value.strftime('%Y%m%d%H%M.%S')))
elif key == 'accounting_id':
cmd.extend(('-A', value))
if email_events:
cmd.extend(('-m', email_events))
# Setup parallel environment.
if 'job_category' in resource_desc:
job_category = resource_desc['job_category']
try:
parallel_environment = self.category_map[job_category]
except KeyError:
msg = 'No mapping for job_category %r' % job_category
self._logger.error(msg)
raise ValueError(msg)
min_cpus = resource_desc.get('min_cpus', 1)
max_cpus = resource_desc.get('max_cpus', min_cpus)
cmd.extend(('-pe', parallel_environment,
'%d-%d' % (min_cpus, max_cpus)))
# Set resource limits.
if 'resource_limits' in resource_desc:
limits = resource_desc['resource_limits']
if 'cpu_time' in limits:
cpu_time = limits['cpu_time']
cmd.extend(('-l', 'h_cpu=%s' % self._timelimit(cpu_time)))
if 'wallclock_time' in limits:
wall_time = limits['wallclock_time']
cmd.extend(('-l', 'h_rt=%s' % self._timelimit(wall_time)))
# Set default command configuration.
if not self.work_dir:
cmd.append('-cwd')
if inp is None:
cmd.extend(('-i', DEV_NULL))
if out is None:
base = os.path.basename(resource_desc['remote_command'])
cmd.extend(('-o', '%s.stdout' % base))
if err is None:
cmd.extend(('-j', 'yes'))
# Add 'escape' clause.
if 'native_specification' in resource_desc:
cmd.extend(resource_desc['native_specification'])
cmd.append(self._fix_path(resource_desc['remote_command']))
if 'args' in resource_desc:
for arg in resource_desc['args']:
cmd.append(self._fix_path(arg))
self._logger.info('%r', ' '.join(cmd))
try:
process = ShellProc(cmd, DEV_NULL, 'qsub.out', STDOUT, env)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-W block=true' `qsub` option is used to wait for job
completion.
Other job resource keys are processed as follows:
========================= ===========================
Resource Key Translation
========================= ===========================
``submit_as_hold`` -h
------------------------- ---------------------------
rerunnable -r y|n
------------------------- ---------------------------
``working_directory`` Handled in generated script
------------------------- ---------------------------
``job_category`` Ignored
------------------------- ---------------------------
``min_cpus`` -l select= `value` :ncpus=1
------------------------- ---------------------------
``max_cpus`` Ignored
------------------------- ---------------------------
``min_phys_memory`` Ignored
------------------------- ---------------------------
email -M `value`
------------------------- ---------------------------
``email_on_started`` -m b
------------------------- ---------------------------
``email_on_terminated`` -m e
------------------------- ---------------------------
``job_name`` -N `value`
------------------------- ---------------------------
``input_path`` Handled in generated script
------------------------- ---------------------------
``output_path`` Handled in generated script
------------------------- ---------------------------
``error_path`` Handled in generated script
------------------------- ---------------------------
``join_files`` Handled in generated script
------------------------- ---------------------------
``reservation_id`` Ignored
------------------------- ---------------------------
``queue_name`` -q `value`
------------------------- ---------------------------
priority -p `value`
------------------------- ---------------------------
``start_time`` -a `value`
------------------------- ---------------------------
``deadline_time`` Ignored
------------------------- ---------------------------
``accounting_id`` -W group_list= `value`
========================= ===========================
Where `value` is the corresponding resource value.
To support a working directory other than HOME or a
PBS-generated scratch directory, a short script is written with
PBS directives in the header. The script will change to the working
directory and then run the command.
If 'working_directory' is not specified, use current server directory.
If 'input_path' is not specified, use ``/dev/null``.
If 'output_path' is not specified, use ``<remote_command>.stdout``.
If 'error_path' is not specified, use stdout.
If 'native_specification' is specified, it is added to the `qsub`
command just before the name of the generated script. If it contains
a ``select`` clause, then that will prevent generation of a ``select``
clause related to 'min_cpus'.
Some resource limits are also handled:
==================== =========================
Resource Key Translation
==================== =========================
``core_file_size`` Ignored
-------------------- -------------------------
``data_seg_size`` Ignored
-------------------- -------------------------
``file_size`` Ignored
-------------------- -------------------------
``open_files`` Ignored
-------------------- -------------------------
``stack_size`` Ignored
-------------------- -------------------------
``virtual_memory`` Ignored
-------------------- -------------------------
``cpu_time`` Ignored
-------------------- -------------------------
``wallclock_time`` -l walltime= `value`
==================== =========================
Output from `qsub` itself is routed to ``qsub.out``.
If the job reports an error, ``qsub.out`` will be appended to either
`error_path`, or if that was not specified, stdout.
"""
self.home_dir = os.path.expanduser('~')
self.work_dir = ''
cmd = list(self._QSUB)
cmd.extend(('-V', '-W', 'block=true', '-j', 'oe'))
if sys.platform == 'win32': # pragma no cover
prefix = 'REM PBS'
cmd.extend(('-C', '"%s"' % prefix))
suffix = '-qsub.bat'
else:
prefix = '#PBS'
cmd.extend(('-S', '/bin/sh'))
suffix = '.qsub'
env = None
inp, out, err = None, None, None
join_files = False
# Set working directory now, for possible path fixing.
try:
value = resource_desc['working_directory']
except KeyError:
pass
else:
self.work_dir = self._fix_path(value)
# Write script to be submitted rather than putting everything on
# 'qsub' command line. We have to do this since otherwise there's
# no way to set an execution directory or input path.
base = None
if 'job_name' in resource_desc:
base = self._jobname(resource_desc['job_name'])
if not base:
base = os.path.basename(resource_desc['remote_command'])
script_name = '%s%s' % (base, suffix)
native_specification = resource_desc.get('native_specification', [])
with open(script_name, 'w') as script:
if sys.platform == 'win32': # pragma no cover
script.write('@echo off\n')
else:
script.write('#!/bin/sh\n')
# PBS (at least at NAS) requires 'group_list' be set.
if 'accounting_id' in resource_desc:
accounting_id = resource_desc['accounting_id']
else:
accounting_id = self.accounting_id
script.write('%s -W group_list=%s\n' %
(prefix, accounting_id.strip()))
# Process description in fixed, repeatable order.
keys = ('submit_as_hold', 'rerunnable', 'job_environment',
'min_cpus', 'email', 'email_on_started',
'email_on_terminated', 'job_name', 'input_path',
'output_path', 'error_path', 'join_files', 'queue_name',
'priority', 'start_time')
email_events = ''
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == 'submit_as_hold':
if value:
script.write('%s -h\n' % prefix)
elif key == 'rerunnable':
script.write('%s -r %s\n' %
(prefix, 'y' if value else 'n'))
elif key == 'job_environment':
env = value
elif key == 'min_cpus':
# Only write select clause if not in 'native_specification'.
for arg in native_specification:
if 'select' in arg:
break
else:
script.write('%s -l select=%d:ncpus=1\n' %
(prefix, value))
elif key == 'email':
script.write('%s -M %s\n' % (prefix, ','.join(value)))
elif key == 'email_on_started':
email_events += 'b'
elif key == 'email_on_terminated':
email_events += 'e'
elif key == 'job_name':
value = value or base
script.write('%s -N %s\n' % (prefix, self._jobname(value)))
elif key == 'input_path':
inp = value
elif key == 'output_path':
out = value
elif key == 'error_path':
err = value
elif key == 'join_files':
join_files = value
elif key == 'queue_name':
script.write('%s -q %s\n' % (prefix, value))
elif key == 'priority':
script.write('%s -p %d\n' % (prefix, value))
elif key == 'start_time':
script.write('%s -a %s\n' %
(prefix, value.strftime('%Y%m%d%H%M.%S')))
if email_events:
script.write('%s -m %s\n' % (prefix, email_events))
# Set resource limits.
if 'resource_limits' in resource_desc:
limits = resource_desc['resource_limits']
if 'wallclock_time' in limits:
wall_time = limits['wallclock_time']
script.write('%s -l walltime=%s\n' %
(prefix, self._timelimit(wall_time)))
# Have script move to work directory relative to
# home directory on execution host.
home = os.path.realpath(os.path.expanduser('~'))
work = os.path.realpath(self.work_dir or os.getcwd())
if work.startswith(home):
work = work[len(home) + 1:]
if sys.platform == 'win32': # pragma no cover
script.write('cd %HOMEDRIVE%%HOMEPATH%\n')
else:
script.write('cd $HOME\n')
else:
# This can potentially cause problems...
self._logger.warning('work %r not a descendant of home %r',
work, home)
if ' ' in work:
work = '"%s"' % work
script.write('cd %s\n' % work)
script.write(self._fix_path(resource_desc['remote_command']))
if 'args' in resource_desc:
for arg in resource_desc['args']:
arg = self._fix_path(arg)
if ' ' in arg and arg[0] not in ('"', "'"):
arg = '"%s"' % arg
script.write(' %s' % arg)
script.write(' <%s' % (inp or DEV_NULL))
script.write(' >%s' % (out or '%s.stdout' % base))
if join_files or err is None:
script.write(' 2>&1')
else:
script.write(' 2>%s' % err)
script.write('\n')
if sys.platform != 'win32':
os.chmod(script_name, 0700)
# Add 'escape' clause.
cmd.extend(native_specification)
with open(script_name, 'rU') as inp:
self._logger.debug('%s:', script_name)
for line in inp:
self._logger.debug(' %s', line.rstrip())
# Submit job.
cmd.append(os.path.join('.', script_name))
self._logger.info('%r', ' '.join(cmd))
try:
process = ShellProc(cmd, DEV_NULL, 'qsub.out', STDOUT, env)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
if os.path.exists('qsub.out'):
with open('qsub.out', 'rU') as inp:
self._logger.error('qsub.out:')
for line in inp:
self._logger.error(' %s', line.rstrip())
raise
# Submitted, wait for completion.
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(' returning %s', (return_code, error_msg))
if return_code and os.path.exists('qsub.out'):
if join_files or err is None:
qsub_echo = out or '%s.stdout' % base
else:
qsub_echo = err
with open('qsub.out', 'rU') as inp:
with open(qsub_echo, 'a+') as out:
self._logger.error('qsub.out:')
out.write('===== qsub.out =====\n')
for line in inp:
self._logger.error(' %s', line.rstrip())
out.write(line)
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-sync yes' `qsub` option is used to wait for job completion.
Other job resource keys are processed as follows:
========================= =========================
Resource Key Translation
========================= =========================
``submit_as_hold`` -h
------------------------- -------------------------
rerunnable -r yes|no
------------------------- -------------------------
``working_directory`` -wd `value`
------------------------- -------------------------
``job_category`` Sets parallel environment
------------------------- -------------------------
``min_cpus`` Sets parallel environment
------------------------- -------------------------
``max_cpus`` Sets parallel environment
------------------------- -------------------------
``min_phys_memory`` Ignored
------------------------- -------------------------
email -M `value`
------------------------- -------------------------
``email_on_started`` -m b
------------------------- -------------------------
``email_on_terminated`` -m e
------------------------- -------------------------
``job_name`` -N `value`
------------------------- -------------------------
``input_path`` -i `value`
------------------------- -------------------------
``output_path`` -o `value`
------------------------- -------------------------
``error_path`` -e `value`
------------------------- -------------------------
``join_files`` -j yes|no
------------------------- -------------------------
``reservation_id`` -ar `value`
------------------------- -------------------------
``queue_name`` -q `value`
------------------------- -------------------------
priority -p `value`
------------------------- -------------------------
``start_time`` -a `value`
------------------------- -------------------------
``deadline_time`` Ignored
------------------------- -------------------------
``accounting_id`` -A `value`
========================= =========================
Where `value` is the corresponding resource value.
If 'working_directory' is not specified, add ``-cwd``.
If 'input_path' is not specified, add ``-i /dev/null``.
If 'output_path' is not specified, add ``-o <remote_command>.stdout``.
If 'error_path' is not specified, add ``-j yes``.
If 'native_specification' is specified, it is added to the `qsub`
command just before 'remote_command' and 'args'.
If specified, 'job_category' is used to index into the category
map set up during allocator configuration. The mapped category
name as well as the 'min_cpus' and 'max_cpus' values are used
with the ``-pe`` qsub option.
Some resource limits are also handled:
==================== =========================
Resource Key Translation
==================== =========================
``core_file_size`` Ignored
-------------------- -------------------------
``data_seg_size`` Ignored
-------------------- -------------------------
``file_size`` Ignored
-------------------- -------------------------
``open_files`` Ignored
-------------------- -------------------------
``stack_size`` Ignored
-------------------- -------------------------
``virtual_memory`` Ignored
-------------------- -------------------------
``cpu_time`` -l h_cpu= `value`
-------------------- -------------------------
``wallclock_time`` -l h_rt= `value`
==================== =========================
Output from `qsub` itself is routed to ``qsub.out``.
"""
self.home_dir = os.path.expanduser('~')
self.work_dir = ''
cmd = list(self._QSUB)
cmd.extend(('-V', '-sync', 'yes', '-b', 'yes'))
env = None
inp, out, err = None, None, None
# Set working directory now, for possible path fixing.
try:
value = resource_desc['working_directory']
except KeyError:
pass
else:
self.work_dir = self._fix_path(value)
cmd.extend(('-wd', value))
# Process description in fixed, repeatable order.
keys = ('submit_as_hold', 'rerunnable', 'job_environment', 'email',
'email_on_started', 'email_on_terminated', 'job_name',
'input_path', 'output_path', 'error_path', 'join_files',
'reservation_id', 'queue_name', 'priority', 'start_time',
'accounting_id')
email_events = ''
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == 'submit_as_hold':
if value:
cmd.append('-h')
elif key == 'rerunnable':
cmd.extend(('-r', 'yes' if value else 'no'))
elif key == 'job_environment':
env = value
elif key == 'email':
cmd.extend(('-M', ','.join(value)))
elif key == 'email_on_started':
email_events += 'b'
elif key == 'email_on_terminated':
email_events += 'e'
elif key == 'job_name':
if value:
cmd.extend(('-N', self._jobname(value)))
elif key == 'input_path':
cmd.extend(('-i', self._fix_path(value)))
inp = value
elif key == 'output_path':
cmd.extend(('-o', self._fix_path(value)))
out = value
elif key == 'error_path':
cmd.extend(('-e', self._fix_path(value)))
err = value
elif key == 'join_files':
cmd.extend(('-j', 'yes' if value else 'no'))
if value:
err = 'yes'
elif key == 'reservation_id':
cmd.extend(('-ar', value))
elif key == 'queue_name':
cmd.extend(('-q', value))
elif key == 'priority':
cmd.extend(('-p', str(value)))
elif key == 'start_time':
cmd.extend(('-a', value.strftime('%Y%m%d%H%M.%S')))
elif key == 'accounting_id':
cmd.extend(('-A', value))
if email_events:
cmd.extend(('-m', email_events))
# Setup parallel environment.
if 'job_category' in resource_desc:
job_category = resource_desc['job_category']
try:
parallel_environment = self.category_map[job_category]
except KeyError:
msg = 'No mapping for job_category %r' % job_category
self._logger.error(msg)
raise ValueError(msg)
min_cpus = resource_desc.get('min_cpus', 1)
max_cpus = resource_desc.get('max_cpus', min_cpus)
cmd.extend(
('-pe', parallel_environment, '%d-%d' % (min_cpus, max_cpus)))
# Set resource limits.
if 'resource_limits' in resource_desc:
limits = resource_desc['resource_limits']
if 'cpu_time' in limits:
cpu_time = limits['cpu_time']
cmd.extend(('-l', 'h_cpu=%s' % self._timelimit(cpu_time)))
if 'wallclock_time' in limits:
wall_time = limits['wallclock_time']
cmd.extend(('-l', 'h_rt=%s' % self._timelimit(wall_time)))
# Set default command configuration.
if not self.work_dir:
cmd.append('-cwd')
if inp is None:
cmd.extend(('-i', DEV_NULL))
if out is None:
base = os.path.basename(resource_desc['remote_command'])
cmd.extend(('-o', '%s.stdout' % base))
if err is None:
cmd.extend(('-j', 'yes'))
# Add 'escape' clause.
if 'native_specification' in resource_desc:
cmd.extend(resource_desc['native_specification'])
cmd.append(self._fix_path(resource_desc['remote_command']))
if 'args' in resource_desc:
for arg in resource_desc['args']:
cmd.append(self._fix_path(arg))
self._logger.info('%r', ' '.join(cmd))
try:
process = ShellProc(cmd, DEV_NULL, 'qsub.out', STDOUT, env)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-sync yes' `qsub` option is used to wait for job completion.
Other job resource keys are processed as follows:
========================= =========================
Resource Key Translation
========================= =========================
``submit_as_hold`` -h
------------------------- -------------------------
``rerunnable`` -r yes|no
------------------------- -------------------------
``working_directory`` Ignored
------------------------- -------------------------
``job_category`` Sets parallel environment
------------------------- -------------------------
``min_cpus`` Sets parallel environment
------------------------- -------------------------
``max_cpus`` Sets parallel environment
------------------------- -------------------------
``min_phys_memory`` Ignored
------------------------- -------------------------
``email`` -M `value`
------------------------- -------------------------
``email_on_started`` -m b
------------------------- -------------------------
``email_on_terminated`` -m e
------------------------- -------------------------
``job_name`` -N `value`
------------------------- -------------------------
``input_path`` -i `value`
------------------------- -------------------------
``output_path`` -o `value`
------------------------- -------------------------
``error_path`` -e `value`
------------------------- -------------------------
``join_files`` -j yes|no
------------------------- -------------------------
``reservation_id`` -ar `value`
------------------------- -------------------------
``queue_name`` -q `value`
------------------------- -------------------------
``priority`` -p `value`
------------------------- -------------------------
``start_time`` -a `value`
------------------------- -------------------------
``deadline_time`` Ignored
------------------------- -------------------------
``accounting_id`` -A `value`
========================= =========================
Where `value` is the corresponding resource value.
The 'working_directory' key is ignored since the server has been
started in this directory. ``-cwd`` is used in the `qsub` command.
If 'input_path' is not specified, add ``-i /dev/null``.
If 'output_path' is not specified, add ``-o <remote_command>.stdout``.
If 'error_path' is not specified, add ``-j yes``.
If 'native_specification' is specified, it is added to the `qsub`
command just before 'remote_command' and 'args'.
If specified, 'job_category' is used to index into the category
map set up during allocator configuration. The mapped category
name as well as the 'min_cpus' and 'max_cpus' values are used
with the ``-pe`` qsub option.
Some resource limits are also handled:
==================== =========================
Resource Key Translation
==================== =========================
``core_file_size`` Ignored
-------------------- -------------------------
``data_seg_size`` Ignored
-------------------- -------------------------
``file_size`` Ignored
-------------------- -------------------------
``open_files`` Ignored
-------------------- -------------------------
``stack_size`` Ignored
-------------------- -------------------------
``virtual_memory`` Ignored
-------------------- -------------------------
``cpu_time`` -l h_cpu= `value`
-------------------- -------------------------
``wallclock_time`` -l h_rt= `value`
==================== =========================
Output from `qsub` itself is routed to ``qsub.out``.
"""
self.home_dir = os.path.expanduser("~")
self.work_dir = os.getcwd() # Server started in working directory.
cmd = list(self._QSUB)
cmd.extend(("-V", "-sync", "yes", "-b", "yes", "-cwd"))
env = None
inp, out, err = None, None, None
# Process description in fixed, repeatable order.
keys = (
"submit_as_hold",
"rerunnable",
"job_environment",
"email",
"email_on_started",
"email_on_terminated",
"job_name",
"input_path",
"output_path",
"error_path",
"join_files",
"reservation_id",
"queue_name",
"priority",
"start_time",
"accounting_id",
)
email_events = ""
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == "submit_as_hold":
if value:
cmd.append("-h")
elif key == "rerunnable":
cmd.extend(("-r", "yes" if value else "no"))
elif key == "job_environment":
env = value
elif key == "email":
cmd.extend(("-M", ",".join(value)))
elif key == "email_on_started":
email_events += "b"
elif key == "email_on_terminated":
email_events += "e"
elif key == "job_name":
if value:
cmd.extend(("-N", self._jobname(value)))
elif key == "input_path":
cmd.extend(("-i", self._fix_path(value)))
inp = value
elif key == "output_path":
cmd.extend(("-o", self._fix_path(value)))
out = value
elif key == "error_path":
cmd.extend(("-e", self._fix_path(value)))
err = value
elif key == "join_files":
cmd.extend(("-j", "yes" if value else "no"))
if value:
err = "yes"
elif key == "reservation_id":
cmd.extend(("-ar", value))
elif key == "queue_name":
cmd.extend(("-q", value))
elif key == "priority":
cmd.extend(("-p", str(value)))
elif key == "start_time":
cmd.extend(("-a", value.strftime("%Y%m%d%H%M.%S")))
elif key == "accounting_id":
cmd.extend(("-A", value))
if email_events:
cmd.extend(("-m", email_events))
# Setup parallel environment.
if "job_category" in resource_desc:
job_category = resource_desc["job_category"]
try:
parallel_environment = self.category_map[job_category]
except KeyError:
msg = "No mapping for job_category %r" % job_category
self._logger.error(msg)
raise ValueError(msg)
min_cpus = resource_desc.get("min_cpus", 1)
max_cpus = resource_desc.get("max_cpus", min_cpus)
cmd.extend(("-pe", parallel_environment, "%d-%d" % (min_cpus, max_cpus)))
# Set resource limits.
if "resource_limits" in resource_desc:
limits = resource_desc["resource_limits"]
if "cpu_time" in limits:
cpu_time = limits["cpu_time"]
cmd.extend(("-l", "h_cpu=%s" % self._timelimit(cpu_time)))
if "wallclock_time" in limits:
wall_time = limits["wallclock_time"]
cmd.extend(("-l", "h_rt=%s" % self._timelimit(wall_time)))
# Set default command configuration.
if inp is None:
cmd.extend(("-i", DEV_NULL))
if out is None:
base = os.path.basename(resource_desc["remote_command"])
cmd.extend(("-o", "%s.stdout" % base))
if err is None:
cmd.extend(("-j", "yes"))
# Add 'escape' clause.
if "native_specification" in resource_desc:
cmd.extend(resource_desc["native_specification"])
cmd.append(self._fix_path(resource_desc["remote_command"]))
if "args" in resource_desc:
for arg in resource_desc["args"]:
cmd.append(self._fix_path(arg))
self._logger.info("%r", " ".join(cmd))
try:
process = ShellProc(cmd, DEV_NULL, "qsub.out", STDOUT, env)
except Exception as exc:
self._logger.error("exception creating process: %s", exc)
raise
self._logger.debug(" PID = %d", process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(" returning %s", (return_code, error_msg))
return (return_code, error_msg)
def start_server(authkey='PublicKey',
address=None,
port=0,
prefix='server',
allowed_hosts=None,
allowed_users=None,
allow_shell=False,
allowed_types=None,
timeout=None,
tunnel=False,
resources=None,
log_prefix=None):
"""
Start an :class:`ObjServerFactory` service in a separate process
in the current directory.
authkey: string
Authorization key; must be matched by clients.
address: string
IPv4 address, hostname, or pipe name.
Default is the host's default IPv4 address.
port: int
Server port (default of 0 implies next available port).
Note that ports below 1024 typically require special privileges.
If port is negative, then a local pipe is used for communication.
prefix: string
Prefix for server config file and stdout/stderr file.
allowed_hosts: list(string)
Host address patterns to check against. Required if `port` >= 0.
Ignored if `allowed_users` is specified.
allowed_users: dict
Dictionary of users and corresponding public keys allowed access.
If None, *any* user may access. If empty, no user may access.
The host portions of user strings are used for address patterns.
allow_shell: bool
If True, :meth:`execute_command` and :meth:`load_model` are allowed.
Use with caution!
allowed_types: list(string)
Names of types which may be created. If None, then allow types listed
by :meth:`get_available_types`. If empty, no types are allowed.
timeout: int
Seconds to wait for server to start. Note that public key generation
can take a while. The default value of None will use an internally
computed value based on host type (and for Windows, the availability
of pyWin32).
tunnel: bool
If True, report host IP address but listen for connections from a
local SSH tunnel.
resources: string
Filename for resource configuration.
log_prefix: string
Name used to identify remote remote logging messages from server.
Implies that the local process will be receiving the messages.
Returns ``(server_proc, config_filename)``.
"""
if timeout is None:
if sys.platform == 'win32' and not HAVE_PYWIN32: # pragma no cover
timeout = 120
else:
timeout = 30
server_key = prefix + '.key'
server_cfg = prefix + '.cfg'
server_out = prefix + '.out'
for path in (server_cfg, server_out):
if os.path.exists(path):
os.remove(path)
with open(server_key, 'w') as out:
out.write('%s\n' % authkey)
factory_path = pkg_resources.resource_filename('openmdao.main',
'objserverfactory.py')
args = ['python', factory_path, '--port', str(port), '--prefix', prefix]
if address is not None:
args.extend(['--address', address])
if tunnel:
args.append('--tunnel')
if resources is not None:
args.append('--resources')
args.append(resources)
if allowed_users is not None:
write_authorized_keys(allowed_users, 'users.allow',
logging.getLogger())
args.extend(['--users', 'users.allow'])
else:
args.append('--allow-public')
if port >= 0:
if allowed_hosts is None:
allowed_hosts = [socket.gethostbyname(socket.gethostname())]
if allowed_hosts[0].startswith('127.') and \
'127.0.0.1' not in allowed_hosts:
allowed_hosts.append('127.0.0.1')
with open('hosts.allow', 'w') as out:
for pattern in allowed_hosts:
out.write('%s\n' % pattern)
if sys.platform != 'win32' or HAVE_PYWIN32:
make_private('hosts.allow')
else: # pragma no cover
logging.warning("Can't make hosts.allow private")
if allow_shell:
args.append('--allow-shell')
if allowed_types is not None:
with open('types.allow', 'w') as out:
for typname in allowed_types:
out.write('%s\n' % typname)
if sys.platform != 'win32' or HAVE_PYWIN32:
make_private('types.allow')
else: # pragma no cover
logging.warning("Can't make types.allow private")
args.extend(['--types', 'types.allow'])
if log_prefix is not None:
log_host = socket.gethostname()
log_port = logging_port(log_host, log_host)
args.extend(['--log-host', log_host, '--log-port', str(log_port)])
if log_prefix: # Could be null (for default).
args.extend(['--log-prefix', log_prefix])
proc = ShellProc(args, stdout=server_out, stderr=STDOUT)
try:
# Wait for valid server_cfg file.
retry = 0
while (not os.path.exists(server_cfg)) or \
(os.path.getsize(server_cfg) == 0):
return_code = proc.poll()
if return_code:
error_msg = proc.error_message(return_code)
raise RuntimeError('Server startup failed %s' % error_msg)
retry += 1
if retry < 10 * timeout:
time.sleep(.1)
# Hard to cause a startup timeout.
else: # pragma no cover
proc.terminate(timeout)
raise RuntimeError('Server startup timeout')
return (proc, server_cfg)
finally:
if os.path.exists(server_key):
os.remove(server_key)
class ExternalCode(ComponentWithDerivatives):
""" Run an external code as a component. """
PIPE = subprocess.PIPE
STDOUT = subprocess.STDOUT
# pylint: disable-msg=E1101
command = Str('',
desc='The command to be executed.')
env_vars = Dict({}, iotype='in',
desc='Environment variables required by the command.')
resources = Dict({}, iotype='in',
desc='Resources required to run this component.')
poll_delay = Float(0., low=0., units='s', iotype='in',
desc='Delay between polling for command completion.'
' A value of zero will use an internally computed'
' default.')
timeout = Float(0., low=0., iotype='in', units='s',
desc='Maximum time to wait for command completion.'
' A value of zero implies an infinite wait.')
timed_out = Bool(False, iotype='out',
desc='True if the command timed-out.')
return_code = Int(0, iotype='out',
desc='Return code from the command.')
def __init__(self, *args, **kwargs):
super(ExternalCode, self).__init__(*args, **kwargs)
self.stdin = None
self.stdout = None
self.stderr = "error.out"
self._process = None
self._server = None
# This gets used by remote server.
def get_access_controller(self): #pragma no cover
""" Return :class:`AccessController` for this object. """
return _AccessController()
@rbac(('owner', 'user'))
def set(self, path, value, index=None, src=None, force=False):
""" Don't allow setting of 'command' by remote client. """
if path in ('command', 'get_access_controller') and remote_access():
self.raise_exception('%r may not be set() remotely' % path,
RuntimeError)
return super(ExternalCode, self).set(path, value, index, src, force)
def execute(self):
"""
Runs the specified command.
First removes existing output (but not in/out) files.
Then if `resources` have been specified, an appropriate server
is allocated and the command is run on that server.
Otherwise the command is run locally.
"""
self.return_code = -12345678
self.timed_out = False
for metadata in self.external_files:
if metadata.get('output', False) and \
not metadata.get('input', False):
for path in glob.glob(metadata.path):
if os.path.exists(path):
os.remove(path)
if not self.command:
self.raise_exception('Null command line', ValueError)
return_code = None
error_msg = ''
try:
if self.resources:
return_code, error_msg = self._execute_remote()
else:
return_code, error_msg = self._execute_local()
if return_code is None:
if self._stop:
self.raise_exception('Run stopped', RunStopped)
else:
self.timed_out = True
self.raise_exception('Timed out', RunInterrupted)
elif return_code:
if isinstance(self.stderr, str):
stderrfile = open(self.stderr, 'r')
error_desc = stderrfile.read()
stderrfile.close()
err_fragment = "\nError Output:\n%s" % error_desc
else:
err_fragment = error_msg
self.raise_exception('return_code = %d%s' \
% (return_code, err_fragment), RuntimeError)
finally:
self.return_code = -999999 if return_code is None else return_code
def _execute_local(self):
""" Run command. """
self._logger.info("executing '%s'...", self.command)
start_time = time.time()
self._process = \
ShellProc(self.command, self.stdin, self.stdout, self.stderr,
self.env_vars)
self._logger.debug('PID = %d', self._process.pid)
try:
return_code, error_msg = \
self._process.wait(self.poll_delay, self.timeout)
finally:
self._process.close_files()
self._process = None
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %.1f sec.', et)
return (return_code, error_msg)
def _execute_remote(self):
"""
Allocate a server based on required resources, send inputs,
run command, and retrieve results.
"""
# Allocate server.
self._server, server_info = RAM.allocate(self.resources)
if self._server is None:
self.raise_exception('Server allocation failed :-(', RuntimeError)
return_code = -88888888
error_msg = ''
try:
# Send inputs.
patterns = []
for metadata in self.external_files:
if metadata.get('input', False):
patterns.append(metadata.path)
if patterns:
self._send_inputs(patterns)
else:
self._logger.debug("No input metadata paths")
# Run command.
self._logger.info("executing '%s'...", self.command)
start_time = time.time()
return_code, error_msg = \
self._server.execute_command(self.command, self.stdin,
self.stdout, self.stderr,
self.env_vars, self.poll_delay,
self.timeout)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %f sec.', et)
# Retrieve results.
patterns = []
for metadata in self.external_files:
if metadata.get('output', False):
patterns.append(metadata.path)
if patterns:
self._retrieve_results(patterns)
else:
self._logger.debug("No output metadata paths")
finally:
RAM.release(self._server)
self._server = None
return (return_code, error_msg)
def _send_inputs(self, patterns):
""" Sends input files matching `patterns`. """
self._logger.info('sending inputs...')
start_time = time.time()
filename = 'inputs.zip'
pfiles, pbytes = pack_zipfile(patterns, filename, self._logger)
try:
filexfer(None, filename, self._server, filename, 'b')
ufiles, ubytes = self._server.unpack_zipfile(filename)
finally:
os.remove(filename)
# Difficult to force file transfer error.
if ufiles != pfiles or ubytes != pbytes: #pragma no cover
msg = 'Inputs xfer error: %d:%d vs. %d:%d' \
% (ufiles, ubytes, pfiles, pbytes)
self.raise_exception(msg, RuntimeError)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %f sec.', et)
def _retrieve_results(self, patterns):
""" Retrieves result files matching `patterns`. """
self._logger.info('retrieving results...')
start_time = time.time()
filename = 'outputs.zip'
pfiles, pbytes = self._server.pack_zipfile(tuple(patterns), filename)
try:
filexfer(self._server, filename, None, filename, 'b')
ufiles, ubytes = unpack_zipfile(filename, self._logger)
finally:
os.remove(filename)
# Difficult to force file transfer error.
if ufiles != pfiles or ubytes != pbytes: #pragma no cover
msg = 'Results xfer error: %d:%d vs. %d:%d' \
% (ufiles, ubytes, pfiles, pbytes)
self.raise_exception(msg, RuntimeError)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %f sec.', et)
def stop(self):
""" Stop the external code. """
self._stop = True
if self._process:
self._process.terminate()
def copy_inputs(self, inputs_dir, patterns):
"""
Copy inputs from `inputs_dir` that match `patterns`.
inputs_dir: string
Directory to copy files from. Relative paths are evaluated from
the component's execution directory.
patterns: list or string
One or more :mod:`glob` patterns to match against.
This can be useful for resetting problem state.
"""
self._logger.info('copying initial inputs from %s...', inputs_dir)
with self.dir_context:
if not os.path.exists(inputs_dir):
self.raise_exception("inputs_dir '%s' does not exist" \
% inputs_dir, RuntimeError)
self._copy(inputs_dir, patterns)
def copy_results(self, results_dir, patterns):
"""
Copy files from `results_dir` that match `patterns`.
results_dir: string
Directory to copy files from. Relative paths are evaluated from
the component's execution directory.
patterns: list or string
One or more :mod:`glob` patterns to match against.
This can be useful for workflow debugging when the external
code takes a long time to execute.
"""
self._logger.info('copying precomputed results from %s...', results_dir)
with self.dir_context:
if not os.path.exists(results_dir):
self.raise_exception("results_dir '%s' does not exist" \
% results_dir, RuntimeError)
self._copy(results_dir, patterns)
def _copy(self, directory, patterns):
"""
Copy files from `directory` that match `patterns`
to the current directory and ensure they are writable.
directory: string
Directory to copy files from.
patterns: list or string
One or more :mod:`glob` patterns to match against.
"""
if isinstance(patterns, basestring):
patterns = [patterns]
for pattern in patterns:
pattern = os.path.join(directory, pattern)
for src_path in sorted(glob.glob(pattern)):
dst_path = os.path.basename(src_path)
self._logger.debug(' %s', src_path)
shutil.copy(src_path, dst_path)
# Ensure writable.
mode = os.stat(dst_path).st_mode
mode |= stat.S_IWUSR
os.chmod(dst_path, mode)
def execute_command(self, resource_desc):
"""
Run command described by `resource_desc` in a subprocess if this
server's `allow_shell` attribute is True.
resource_desc: dict
Contains job description.
The current environment, along with any 'job_environment' specification,
is in effect while running 'remote_command'.
If 'input_path' is not specified, ``/dev/null`` or ``nul:`` is used.
If 'output_path' is not specified, ``<remote_command>.stdout`` is used.
If neither 'error_path' nor 'join_files' are specified,
``<remote_command>.stderr`` is used.
If specified in the 'resource_limits' dictionary, 'wallclock_time' is
used as a timeout.
All other queuing resource keys are ignored.
The ``HOME_DIRECTORY`` and ``WORKING_DIRECTORY`` placeholders are
ignored.
"""
try:
job_name = resource_desc['job_name']
except KeyError:
job_name = ''
command = resource_desc['remote_command']
self._check_path(command, 'execute_command')
base = os.path.basename(command)
command = [command]
if 'args' in resource_desc:
command.extend(resource_desc['args'])
self._logger.debug('execute_command %s %r', job_name, command)
if not self._allow_shell:
self._logger.error('attempt to execute %r by %r', command,
get_credentials().user)
raise RuntimeError('shell access is not allowed by this server')
env_vars = resource_desc.get('job_environment')
try:
stdin = resource_desc['input_path']
self._check_path(stdin, 'execute_command')
except KeyError:
stdin = DEV_NULL
try:
stdout = resource_desc['output_path']
self._check_path(stdout, 'execute_command')
except KeyError:
stdout = base+'.stdout'
try:
stderr = resource_desc['error_path']
self._check_path(stderr, 'execute_command')
except KeyError:
try:
join_files = resource_desc['join_files']
except KeyError:
stderr = base+'.stderr'
else:
stderr = STDOUT if join_files else base+'.stderr'
limits = resource_desc.get('resource_limits', {})
timeout = limits.get('wallclock_time', 0)
poll_delay = 1
try:
process = ShellProc(command, stdin, stdout, stderr, env_vars)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(poll_delay, timeout)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-W block=true' `qsub` option is used to wait for job
completion.
Other job resource keys are processed as follows:
========================= ===========================
Resource Key Translation
========================= ===========================
submit_as_hold -h
------------------------- ---------------------------
rerunnable -r y|n
------------------------- ---------------------------
working_directory Handled in generated script
------------------------- ---------------------------
job_category Ignored
------------------------- ---------------------------
min_cpus -l select= `value` :ncpus=1
------------------------- ---------------------------
max_cpus Ignored
------------------------- ---------------------------
min_phys_memory Ignored
------------------------- ---------------------------
email -M `value`
------------------------- ---------------------------
email_on_started -m b
------------------------- ---------------------------
email_on_terminated -m e
------------------------- ---------------------------
job_name -N `value`
------------------------- ---------------------------
input_path Handled in generated script
------------------------- ---------------------------
output_path Handled in generated script
------------------------- ---------------------------
error_path Handled in generated script
------------------------- ---------------------------
join_files Handled in generated script
------------------------- ---------------------------
reservation_id Ignored
------------------------- ---------------------------
queue_name -q `value`
------------------------- ---------------------------
priority -p `value`
------------------------- ---------------------------
start_time -a `value`
------------------------- ---------------------------
deadline_time Ignored
------------------------- ---------------------------
accounting_id -W group_list= `value`
========================= ===========================
Where `value` is the corresponding resource value.
In order to support a working directory other than HOME or a
PBS-generated scratch directory, a short script is written with
PBS directives in the header. The script will change to the working
directory and then run the command.
If 'working_directory' is not specified, use current server directory.
If 'input_path' is not specified, use ``/dev/null``.
If 'output_path' is not specified, use ``<remote_command>.stdout``.
If 'error_path' is not specified, use stdout.
If 'native_specification' is specified, it is added to the `qsub`
command just before the name of the generated script. If it contains
a ``select`` clause, then that will prevent generation of a ``select``
clause related to 'min_cpus'.
Some resource limits are also handled:
==================== =========================
Resource Key Translation
==================== =========================
core_file_size Ignored
-------------------- -------------------------
data_seg_size Ignored
-------------------- -------------------------
file_size Ignored
-------------------- -------------------------
open_files Ignored
-------------------- -------------------------
stack_size Ignored
-------------------- -------------------------
virtual_memory Ignored
-------------------- -------------------------
cpu_time Ignored
-------------------- -------------------------
wallclock_time -l walltime= `value`
==================== =========================
Output from `qsub` itself is routed to ``qsub.out``.
"""
self.home_dir = os.path.expanduser('~')
self.work_dir = ''
cmd = list(self._QSUB)
cmd.extend(('-V', '-W', 'block=true', '-j', 'oe'))
if sys.platform == 'win32': # pragma no cover
prefix = 'REM PBS'
cmd.extend(('-C', '"%s"' % prefix))
suffix = '-qsub.bat'
else:
prefix = '#PBS'
cmd.extend(('-S', '/bin/sh'))
suffix = '.qsub'
env = None
inp, out, err = None, None, None
join_files = False
# Set working directory now, for possible path fixing.
try:
value = resource_desc['working_directory']
except KeyError:
pass
else:
self.work_dir = self._fix_path(value)
# Write script to be submitted rather than putting everything on
# 'qsub' command line. We have to do this since otherwise there's
# no way to set an execution directory or input path.
if 'job_name' in resource_desc:
base = self._jobname(resource_desc['job_name'])
else:
base = os.path.basename(resource_desc['remote_command'])
script_name = '%s%s' % (base, suffix)
native_specification = resource_desc.get('native_specification', [])
with open(script_name, 'w') as script:
if sys.platform == 'win32': # pragma no cover
script.write('@echo off\n')
else:
script.write('#!/bin/sh\n')
# PBS (at least at NAS) requires 'group_list' be set.
if 'accounting_id' in resource_desc:
accounting_id = resource_desc['accounting_id']
else:
accounting_id = self.accounting_id
script.write('%s -W group_list=%s\n'
% (prefix, accounting_id.strip()))
# Process description in fixed, repeatable order.
keys = ('submit_as_hold',
'rerunnable',
'job_environment',
'min_cpus',
'email',
'email_on_started',
'email_on_terminated',
'job_name',
'input_path',
'output_path',
'error_path',
'join_files',
'queue_name',
'priority',
'start_time')
email_events = ''
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == 'submit_as_hold':
if value:
script.write('%s -h\n' % prefix)
elif key == 'rerunnable':
script.write('%s -r %s\n' % (prefix, 'y' if value else 'n'))
elif key == 'job_environment':
env = value
elif key == 'min_cpus':
# Only write select clause if not in 'native_specification'.
for arg in native_specification:
if 'select' in arg:
break
else:
script.write('%s -l select=%d:ncpus=1\n' % (prefix, value))
elif key == 'email':
script.write('%s -M %s\n' % (prefix, ','.join(value)))
elif key == 'email_on_started':
email_events += 'b'
elif key == 'email_on_terminated':
email_events += 'e'
elif key == 'job_name':
script.write('%s -N %s\n' % (prefix, self._jobname(value)))
elif key == 'input_path':
inp = value
elif key == 'output_path':
out = value
elif key == 'error_path':
err = value
elif key == 'join_files':
join_files = value
elif key == 'queue_name':
script.write('%s -q %s\n' % (prefix, value))
elif key == 'priority':
script.write('%s -p %d\n' % (prefix, value))
elif key == 'start_time':
script.write('%s -a %s\n'
% (prefix, value.strftime('%Y%m%d%H%M.%S')))
if email_events:
script.write('%s -m %s\n' % (prefix, email_events))
# Set resource limits.
if 'resource_limits' in resource_desc:
limits = resource_desc['resource_limits']
if 'wallclock_time' in limits:
wall_time = limits['wallclock_time']
script.write('%s -l walltime=%s\n'
% (prefix, self._timelimit(wall_time)))
# Have script move to work directory relative to
# home directory on execution host.
home = os.path.realpath(os.path.expanduser('~'))
work = os.path.realpath(self.work_dir or os.getcwd())
if work.startswith(home):
work = work[len(home)+1:]
if sys.platform == 'win32': # pragma no cover
script.write('cd %HOMEDRIVE%%HOMEPATH%\n')
else:
script.write('cd $HOME\n')
else:
# This can potentially cause problems...
self._logger.warning('work %r not a descendant of home %r',
work, home)
script.write('cd %s\n' % work)
script.write(self._fix_path(resource_desc['remote_command']))
if 'args' in resource_desc:
for arg in resource_desc['args']:
script.write(' %s' % self._fix_path(arg))
script.write(' <%s' % (inp or DEV_NULL))
script.write(' >%s' % (out or '%s.stdout' % base))
if join_files or err is None:
script.write(' 2>&1')
else:
script.write(' 2>%s' % err)
script.write('\n')
if sys.platform != 'win32':
os.chmod(script_name, 0700)
# Add 'escape' clause.
cmd.extend(native_specification)
cmd.append(os.path.join('.', script_name))
self._logger.info('%r', ' '.join(cmd))
try:
process = ShellProc(cmd, DEV_NULL, 'qsub.out', STDOUT, env)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
def execute_command(self, resource_desc):
"""
Submit command based on `resource_desc`.
resource_desc: dict
Description of command and required resources.
The '-V' `qsub` option is always used to export the current environment
to the job. This environment is first updated with any 'job_environment'
data. The '-sync yes' `qsub` option is used to wait for job completion.
Other job resource keys are processed as follows:
========================= ====================
Resource Key Translation
========================= ====================
job_name -N `value`
------------------------- --------------------
working_directory -wd `value`
------------------------- --------------------
parallel_environment -pe `value` `n_cpus`
------------------------- --------------------
input_path -i `value`
------------------------- --------------------
output_path -o `value`
------------------------- --------------------
error_path -e `value`
------------------------- --------------------
join_files -j yes|no
------------------------- --------------------
email -M `value`
------------------------- --------------------
block_email -m n
------------------------- --------------------
email_events -m `value`
------------------------- --------------------
start_time -a `value`
------------------------- --------------------
deadline_time Not supported
------------------------- --------------------
hard_wallclock_time_limit -l h_rt= `value`
------------------------- --------------------
soft_wallclock_time_limit -l s_rt= `value`
------------------------- --------------------
hard_run_duration_limit -l h_cpu= `value`
------------------------- --------------------
soft_run_duration_limit -l s_cpu= `value`
------------------------- --------------------
job_category Not supported
========================= ====================
Where `value` is the corresponding resource value and
`n_cpus` is the value of the 'n_cpus' resource, or 1.
If 'working_directory' is not specified, add ``-cwd``.
If 'input_path' is not specified, add ``-i /dev/null``.
If 'output_path' is not specified, add ``-o <remote_command>.stdout``.
If 'error_path' is not specified, add ``-j yes``.
If 'native_specification' is specified, it is added to the `qsub`
command just before 'remote_command' and 'args'.
Output from `qsub` itself is routed to ``qsub.out``.
"""
self.home_dir = os.environ['HOME']
self.work_dir = ''
cmd = [self._QSUB, '-V', '-sync', 'yes']
env = None
inp, out, err = None, None, None
# Set working directory now, for possible path fixing.
try:
value = resource_desc['working_directory']
except KeyError:
pass
else:
self.work_dir = self._fix_path(value)
cmd.append('-wd')
cmd.append(value)
# Process description in fixed, repeatable order.
keys = ('job_name',
'job_environment',
'parallel_environment',
'input_path',
'output_path',
'error_path',
'join_files',
'email',
'block_email',
'email_events',
'start_time',
'hard_wallclock_time_limit',
'soft_wallclock_time_limit',
'hard_run_duration_limit',
'soft_run_duration_limit')
for key in keys:
try:
value = resource_desc[key]
except KeyError:
continue
if key == 'job_name':
cmd.append('-N')
cmd.append(value)
elif key == 'job_environment':
env = value
elif key == 'parallel_environment':
n_cpus = resource_desc.get('n_cpus', 1)
cmd.append('-pe')
cmd.append(value)
cmd.append(str(n_cpus))
elif key == 'input_path':
cmd.append('-i')
cmd.append(self._fix_path(value))
inp = value
elif key == 'output_path':
cmd.append('-o')
cmd.append(self._fix_path(value))
out = value
elif key == 'error_path':
cmd.append('-e')
cmd.append(self._fix_path(value))
err = value
elif key == 'join_files':
cmd.append('-j')
cmd.append('yes' if value else 'no')
if value:
err = 'yes'
elif key == 'email':
cmd.append('-M')
cmd.append(','.join(value))
elif key == 'block_email':
if value:
cmd.append('-m')
cmd.append('n')
elif key == 'email_events':
cmd.append('-m')
cmd.append(value)
elif key == 'start_time':
cmd.append('-a')
cmd.append(value) # May need to translate
elif key == 'hard_wallclock_time_limit':
cmd.append('-l')
cmd.append('h_rt=%s' % self._make_time(value))
elif key == 'soft_wallclock_time_limit':
cmd.append('-l')
cmd.append('s_rt=%s' % self._make_time(value))
elif key == 'hard_run_duration_limit':
cmd.append('-l')
cmd.append('h_cpu=%s' % self._make_time(value))
elif key == 'soft_run_duration_limit':
cmd.append('-l')
cmd.append('s_cpu=%s' % self._make_time(value))
if not self.work_dir:
cmd.append('-cwd')
if inp is None:
cmd.append('-i')
cmd.append('/dev/null')
if out is None:
cmd.append('-o')
cmd.append('%s.stdout'
% os.path.basename(resource_desc['remote_command']))
if err is None:
cmd.append('-j')
cmd.append('yes')
if 'native_specification' in resource_desc:
cmd.extend(resource_desc['native_specification'])
cmd.append(self._fix_path(resource_desc['remote_command']))
if 'args' in resource_desc:
for arg in resource_desc['args']:
cmd.append(self._fix_path(arg))
self._logger.info('%r', ' '.join(cmd))
try:
process = ShellProc(cmd, '/dev/null', 'qsub.out', STDOUT, env)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(1)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
class ExternalCode(ComponentWithDerivatives):
"""
Run an external code as a component. The component can be configured to
run the code on a remote server, see :meth:`execute`.
"""
PIPE = subprocess.PIPE
STDOUT = subprocess.STDOUT
# pylint: disable-msg=E1101
command = List(Str, desc='The command to be executed.')
env_vars = Dict({},
iotype='in',
desc='Environment variables required by the command.')
resources = Dict({},
iotype='in',
desc='Resources required to run this component.')
poll_delay = Float(0.,
low=0.,
units='s',
iotype='in',
desc='Delay between polling for command completion.'
' A value of zero will use an internally computed'
' default.')
timeout = Float(0.,
low=0.,
iotype='in',
units='s',
desc='Maximum time to wait for command completion.'
' A value of zero implies an infinite wait.')
timed_out = Bool(False,
iotype='out',
desc='True if the command timed-out.')
return_code = Int(0, iotype='out', desc='Return code from the command.')
def __init__(self, *args, **kwargs):
super(ExternalCode, self).__init__(*args, **kwargs)
self.stdin = None
self.stdout = None
self.stderr = "error.out"
self._process = None
self._server = None
# This gets used by remote server.
def get_access_controller(self): #pragma no cover
""" Return :class:`AccessController` for this object. """
return _AccessController()
@rbac(('owner', 'user'))
def set(self, path, value, index=None, src=None, force=False):
""" Don't allow setting of 'command' by a remote client. """
if path in ('command', 'get_access_controller') and remote_access():
self.raise_exception('%r may not be set() remotely' % path,
RuntimeError)
return super(ExternalCode, self).set(path, value, index, src, force)
def execute(self):
"""
Runs the specified command.
First removes existing output (but not in/out) files.
Then if `resources` have been specified, an appropriate server
is allocated and the command is run on that server.
Otherwise the command is run locally.
When running remotely, the following resources are set:
======================= =====================================
Key Value
======================= =====================================
job_name self.get_pathname()
----------------------- -------------------------------------
remote_command self.command (first item)
----------------------- -------------------------------------
args self.command (2nd through last items)
----------------------- -------------------------------------
job_environment self.env_vars
----------------------- -------------------------------------
input_path self.stdin
----------------------- -------------------------------------
output_path self.stdout
----------------------- -------------------------------------
error_path self.stderr (if != STDOUT)
----------------------- -------------------------------------
join_files If self.stderr == STDOUT
----------------------- -------------------------------------
hard_run_duration_limit self.timeout (if non-zero)
======================= =====================================
.. note::
Input files to be sent to the remote server are defined by
:class:`FileMetadata` entries in the `external_files` list
with `input` True. Similarly, output files to be retrieved
from the remote server are defined by entries with `output`
True.
.. warning::
Any file **not** labelled with `binary` True will undergo
newline translation if the local and remote machines have
different newline representations. Newline translation will
corrupt a file which is binary but hasn't been labelled as
such.
"""
self.return_code = -12345678
self.timed_out = False
for metadata in self.external_files:
if metadata.get('output', False) and \
not metadata.get('input', False):
for path in glob.glob(metadata.path):
if os.path.exists(path):
os.remove(path)
if not self.command:
self.raise_exception('Null command line', ValueError)
return_code = None
error_msg = ''
try:
if self.resources:
return_code, error_msg = self._execute_remote()
else:
return_code, error_msg = self._execute_local()
if return_code is None:
if self._stop:
self.raise_exception('Run stopped', RunStopped)
else:
self.timed_out = True
self.raise_exception('Timed out', RunInterrupted)
elif return_code:
if isinstance(self.stderr, str):
stderrfile = open(self.stderr, 'r')
error_desc = stderrfile.read()
stderrfile.close()
err_fragment = "\nError Output:\n%s" % error_desc
else:
err_fragment = error_msg
self.raise_exception('return_code = %d%s' \
% (return_code, err_fragment), RuntimeError)
finally:
self.return_code = -999999 if return_code is None else return_code
def _execute_local(self):
""" Run command. """
self._logger.info('executing %s...', self.command)
start_time = time.time()
self._process = \
ShellProc(self.command, self.stdin, self.stdout, self.stderr,
self.env_vars)
self._logger.debug('PID = %d', self._process.pid)
try:
return_code, error_msg = \
self._process.wait(self.poll_delay, self.timeout)
finally:
self._process.close_files()
self._process = None
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %.1f sec.', et)
return (return_code, error_msg)
def _execute_remote(self):
"""
Allocate a server based on required resources, send inputs,
run command, and retrieve results.
"""
# Allocate server.
self._server, server_info = RAM.allocate(self.resources)
if self._server is None:
self.raise_exception('Server allocation failed :-(', RuntimeError)
return_code = -88888888
error_msg = ''
try:
# Create resource description for command.
rdesc = self.resources.copy()
rdesc['job_name'] = self.get_pathname()
rdesc['remote_command'] = self.command[0]
if len(self.command) > 1:
rdesc['args'] = self.command[1:]
if self.env_vars:
rdesc['job_environment'] = self.env_vars
if self.stdin:
rdesc['input_path'] = self.stdin
if self.stdout:
rdesc['output_path'] = self.stdout
if self.stderr:
if self.stderr == self.STDOUT:
rdesc['join_files'] = True
else:
rdesc['error_path'] = self.stderr
if self.timeout:
rdesc['hard_run_duration_limit'] = self.timeout
# Send inputs.
patterns = []
textfiles = []
for metadata in self.external_files:
if metadata.get('input', False):
patterns.append(metadata.path)
if not metadata.binary:
textfiles.append(metadata.path)
if patterns:
self._send_inputs(patterns, textfiles)
else:
self._logger.debug('No input metadata paths')
# Run command.
self._logger.info('executing %s...', self.command)
start_time = time.time()
return_code, error_msg = \
self._server.execute_command(rdesc)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %.1f sec.', et)
# Retrieve results.
patterns = []
textfiles = []
for metadata in self.external_files:
if metadata.get('output', False):
patterns.append(metadata.path)
if not metadata.binary:
textfiles.append(metadata.path)
if patterns:
self._retrieve_results(patterns, textfiles)
else:
self._logger.debug('No output metadata paths')
finally:
RAM.release(self._server)
self._server = None
return (return_code, error_msg)
def _send_inputs(self, patterns, textfiles):
""" Sends input files matching `patterns`. """
self._logger.info('sending inputs...')
start_time = time.time()
filename = 'inputs.zip'
pfiles, pbytes = pack_zipfile(patterns, filename, self._logger)
try:
filexfer(None, filename, self._server, filename, 'b')
ufiles, ubytes = self._server.unpack_zipfile(filename,
textfiles=textfiles)
finally:
os.remove(filename)
# Difficult to force file transfer error.
if ufiles != pfiles or ubytes != pbytes: #pragma no cover
msg = 'Inputs xfer error: %d:%d vs. %d:%d' \
% (ufiles, ubytes, pfiles, pbytes)
self.raise_exception(msg, RuntimeError)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %f sec.', et)
def _retrieve_results(self, patterns, textfiles):
""" Retrieves result files matching `patterns`. """
self._logger.info('retrieving results...')
start_time = time.time()
filename = 'outputs.zip'
pfiles, pbytes = self._server.pack_zipfile(patterns, filename)
filexfer(self._server, filename, None, filename, 'b')
# Valid, but empty, file causes unpack_zipfile() problems.
try:
if os.path.getsize(filename) > 0:
ufiles, ubytes = unpack_zipfile(filename,
logger=self._logger,
textfiles=textfiles)
else:
ufiles, ubytes = 0, 0
finally:
os.remove(filename)
# Difficult to force file transfer error.
if ufiles != pfiles or ubytes != pbytes: #pragma no cover
msg = 'Results xfer error: %d:%d vs. %d:%d' \
% (ufiles, ubytes, pfiles, pbytes)
self.raise_exception(msg, RuntimeError)
et = time.time() - start_time
if et >= 60: #pragma no cover
self._logger.info('elapsed time: %f sec.', et)
def stop(self):
""" Stop the external code. """
self._stop = True
if self._process:
self._process.terminate()
def copy_inputs(self, inputs_dir, patterns):
"""
Copy inputs from `inputs_dir` that match `patterns`.
inputs_dir: string
Directory to copy files from. Relative paths are evaluated from
the component's execution directory.
patterns: list or string
One or more :mod:`glob` patterns to match against.
This can be useful for resetting problem state.
"""
self._logger.info('copying initial inputs from %s...', inputs_dir)
with self.dir_context:
if not os.path.exists(inputs_dir):
self.raise_exception("inputs_dir '%s' does not exist" \
% inputs_dir, RuntimeError)
self._copy(inputs_dir, patterns)
def copy_results(self, results_dir, patterns):
"""
Copy files from `results_dir` that match `patterns`.
results_dir: string
Directory to copy files from. Relative paths are evaluated from
the component's execution directory.
patterns: list or string
One or more :mod:`glob` patterns to match against.
This can be useful for workflow debugging when the external
code takes a long time to execute.
"""
self._logger.info('copying precomputed results from %s...',
results_dir)
with self.dir_context:
if not os.path.exists(results_dir):
self.raise_exception("results_dir '%s' does not exist" \
% results_dir, RuntimeError)
self._copy(results_dir, patterns)
def _copy(self, directory, patterns):
"""
Copy files from `directory` that match `patterns`
to the current directory and ensure they are writable.
directory: string
Directory to copy files from.
patterns: list or string
One or more :mod:`glob` patterns to match against.
"""
if isinstance(patterns, basestring):
patterns = [patterns]
for pattern in patterns:
pattern = os.path.join(directory, pattern)
for src_path in sorted(glob.glob(pattern)):
dst_path = os.path.basename(src_path)
self._logger.debug(' %s', src_path)
shutil.copy(src_path, dst_path)
# Ensure writable.
mode = os.stat(dst_path).st_mode
mode |= stat.S_IWUSR
os.chmod(dst_path, mode)
def start_server(authkey='PublicKey', address=None, port=0, prefix='server',
allowed_hosts=None, allowed_users=None, allow_shell=False,
allowed_types=None, timeout=None, tunnel=False,
resources=None, log_prefix=None):
"""
Start an :class:`ObjServerFactory` service in a separate process
in the current directory.
authkey: string
Authorization key, must be matched by clients.
address: string
IPv4 address, hostname, or pipe name.
Default is the host's default IPv4 address.
port: int
Server port (default of 0 implies next available port).
Note that ports below 1024 typically require special privileges.
If port is negative, then a local pipe is used for communication.
prefix: string
Prefix for server config file and stdout/stderr file.
allowed_hosts: list(string)
Host address patterns to check against. Required if `port` >= 0.
Ignored if `allowed_users` is specified.
allowed_users: dict
Dictionary of users and corresponding public keys allowed access.
If None, *any* user may access. If empty, no user may access.
The host portions of user strings are used for address patterns.
allow_shell: bool
If True, :meth:`execute_command` and :meth:`load_model` are allowed.
Use with caution!
allowed_types: list(string)
Names of types which may be created. If None, then allow types listed
by :meth:`get_available_types`. If empty, no types are allowed.
timeout: int
Seconds to wait for server to start. Note that public key generation
can take a while. The default value of None will use an internally
computed value based on host type (and for Windows, the availability
of pyWin32).
tunnel: bool
If True, report host IP address but listen for connections from a
local SSH tunnel.
resources: string
Filename for resource configuration.
log_prefix: string
Name used to identify remote remote logging messages from server.
Implies that the local process will be receiving the messages.
Returns ``(server_proc, config_filename)``.
"""
if timeout is None:
if sys.platform == 'win32' and not HAVE_PYWIN32: #pragma no cover
timeout = 120
else:
timeout = 30
server_key = prefix+'.key'
server_cfg = prefix+'.cfg'
server_out = prefix+'.out'
for path in (server_cfg, server_out):
if os.path.exists(path):
os.remove(path)
with open(server_key, 'w') as out:
out.write('%s\n' % authkey)
factory_path = pkg_resources.resource_filename('openmdao.main',
'objserverfactory.py')
args = ['python', factory_path, '--port', str(port), '--prefix', prefix]
if address is not None:
args.extend(['--address', address])
if tunnel:
args.append('--tunnel')
if resources is not None:
args.append('--resources')
args.append(resources)
if allowed_users is not None:
write_authorized_keys(allowed_users, 'users.allow', logging.getLogger())
args.extend(['--users', 'users.allow'])
else:
args.append('--allow-public')
if port >= 0:
if allowed_hosts is None:
allowed_hosts = [socket.gethostbyname(socket.gethostname())]
if allowed_hosts[0].startswith('127.') and \
'127.0.0.1' not in allowed_hosts:
allowed_hosts.append('127.0.0.1')
with open('hosts.allow', 'w') as out:
for pattern in allowed_hosts:
out.write('%s\n' % pattern)
if sys.platform != 'win32' or HAVE_PYWIN32:
make_private('hosts.allow')
else: #pragma no cover
logging.warning("Can't make hosts.allow private")
if allow_shell:
args.append('--allow-shell')
if allowed_types is not None:
with open('types.allow', 'w') as out:
for typname in allowed_types:
out.write('%s\n' % typname)
if sys.platform != 'win32' or HAVE_PYWIN32:
make_private('types.allow')
else: #pragma no cover
logging.warning("Can't make types.allow private")
args.extend(['--types', 'types.allow'])
if log_prefix is not None:
log_host = socket.gethostname()
log_port = logging_port(log_host, log_host)
args.extend(['--log-host', log_host, '--log-port', str(log_port)])
if log_prefix: # Could be null (for default).
args.extend(['--log-prefix', log_prefix])
proc = ShellProc(args, stdout=server_out, stderr=STDOUT)
try:
# Wait for valid server_cfg file.
retry = 0
while (not os.path.exists(server_cfg)) or \
(os.path.getsize(server_cfg) == 0):
return_code = proc.poll()
if return_code:
error_msg = proc.error_message(return_code)
raise RuntimeError('Server startup failed %s' % error_msg)
retry += 1
if retry < 10*timeout:
time.sleep(.1)
# Hard to cause a startup timeout.
else: #pragma no cover
proc.terminate(timeout)
raise RuntimeError('Server startup timeout')
return (proc, server_cfg)
finally:
if os.path.exists(server_key):
os.remove(server_key)
def execute_command(self, resource_desc):
"""
Run command described by `resource_desc` in a subprocess if this
server's `allow_shell` attribute is True.
resource_desc: dict
Contains job description.
The current environment, along with any 'job_environment' specification,
is in effect while running 'remote_command'.
If 'input_path' is not specified, ``/dev/null`` or ``nul:`` is used.
If 'output_path' is not specified, ``<remote_command>.stdout`` is used.
If neither 'error_path' nor 'join_files' are specified,
``<remote_command>.stderr`` is used.
If specified in the 'resource_limits' dictionary, 'wallclock_time' is
used as a timeout.
All other queuing resource keys are ignored.
The ``HOME_DIRECTORY`` and ``WORKING_DIRECTORY`` placeholders are
ignored.
"""
try:
job_name = resource_desc['job_name']
except KeyError:
job_name = ''
command = resource_desc['remote_command']
self._check_path(command, 'execute_command')
base = os.path.basename(command)
command = [command]
if 'args' in resource_desc:
command.extend(resource_desc['args'])
self._logger.debug('execute_command %s %r', job_name, command)
if not self._allow_shell:
self._logger.error('attempt to execute %r by %r', command,
get_credentials().user)
raise RuntimeError('shell access is not allowed by this server')
env_vars = resource_desc.get('job_environment')
try:
stdin = resource_desc['input_path']
self._check_path(stdin, 'execute_command')
except KeyError:
stdin = DEV_NULL
try:
stdout = resource_desc['output_path']
self._check_path(stdout, 'execute_command')
except KeyError:
stdout = base + '.stdout'
try:
stderr = resource_desc['error_path']
self._check_path(stderr, 'execute_command')
except KeyError:
try:
join_files = resource_desc['join_files']
except KeyError:
stderr = base + '.stderr'
else:
stderr = STDOUT if join_files else base + '.stderr'
limits = resource_desc.get('resource_limits', {})
timeout = limits.get('wallclock_time', 0)
poll_delay = 1
try:
process = ShellProc(command, stdin, stdout, stderr, env_vars)
except Exception as exc:
self._logger.error('exception creating process: %s', exc)
raise
self._logger.debug(' PID = %d', process.pid)
return_code, error_msg = process.wait(poll_delay, timeout)
self._logger.debug(' returning %s', (return_code, error_msg))
return (return_code, error_msg)
