Source code for aiida.schedulers.plugins.direct

# -*- coding: utf-8 -*-
###########################################################################
# Copyright (c), The AiiDA team. All rights reserved.                     #
# This file is part of the AiiDA code.                                    #
#                                                                         #
# The code is hosted on GitHub at https://github.com/aiidateam/aiida-core #
# For further information on the license, see the LICENSE.txt file        #
# For further information please visit http://www.aiida.net               #
###########################################################################
"""
Plugin for direct execution.
"""

from aiida.common.escaping import escape_for_bash
import aiida.schedulers
from aiida.schedulers import SchedulerError
from aiida.schedulers.datastructures import JobInfo, JobState, NodeNumberJobResource

## From the ps man page on Mac OS X 10.12
#     state     The state is given by a sequence of characters, for example,
#               ``RWNA''.  The first character indicates the run state of the
#               process:
#
#               I       Marks a process that is idle (sleeping for longer than
#                       about 20 seconds).
#               R       Marks a runnable process.
#               S       Marks a process that is sleeping for less than about 20
#                       seconds.
#               T       Marks a stopped process.
#               U       Marks a process in uninterruptible wait.
#               Z       Marks a dead process (a ``zombie'').

# From the man page of ps on Ubuntu 14.04:
#       Here are the different values that the s, stat and state output
#       specifiers (header "STAT" or "S") will display to describe the state of
#       a process:
#
#               D    uninterruptible sleep (usually IO)
#               R    running or runnable (on run queue)
#               S    interruptible sleep (waiting for an event to complete)
#               T    stopped, either by a job control signal or because it is
#                    being traced
#               W    paging (not valid since the 2.6.xx kernel)
#               X    dead (should never be seen)
#               Z    defunct ("zombie") process, terminated but not reaped by
#                    its parent

_MAP_STATUS_PS = {
    'D': JobState.RUNNING,
    'I': JobState.RUNNING,
    'R': JobState.RUNNING,
    'S': JobState.RUNNING,
    'T': JobState.SUSPENDED,
    'U': JobState.RUNNING,
    'W': JobState.RUNNING,
    'X': JobState.DONE,
    'Z': JobState.DONE,
    # Not sure about these three, I comment them out (they used to be in
    # here, but they don't appear neither on ubuntu nor on Mac)
    #    'F': JobState.DONE,
    #    'H': JobState.QUEUED_HELD,
    #    'Q': JobState.QUEUED,
}


[docs]class DirectJobResource(NodeNumberJobResource): """ An implementation of JobResource for the direct excution bypassing schedulers. """
[docs] @classmethod def accepts_default_memory_per_machine(cls): """Return True if this subclass accepts a `default_memory_per_machine` key, False otherwise.""" return False
[docs]class DirectScheduler(aiida.schedulers.Scheduler): """ Support for the direct execution bypassing schedulers. """ _logger = aiida.schedulers.Scheduler._logger.getChild('direct') # Query only by list of jobs and not by user _features = { 'can_query_by_user': True, } # The class to be used for the job resource. _job_resource_class = DirectJobResource
[docs] def _get_joblist_command(self, jobs=None, user=None): """ The command to report full information on existing jobs. TODO: in the case of job arrays, decide what to do (i.e., if we want to pass the -t options to list each subjob). """ # Using subprocess.Popen with `start_new_session=True` (as done in both local and ssh transport) results in # processes without a controlling terminal. # The -x option tells ps to include processes which do not have a controlling terminal, which would not be # listed otherwise (leading the direct scheduler to conclude that the process already completed). command = 'ps -xo pid,stat,user,time' if jobs: if isinstance(jobs, str): command += f' {escape_for_bash(jobs)}' else: try: command += f" {' '.join(escape_for_bash(job) for job in jobs if job)}" except TypeError: raise TypeError("If provided, the 'jobs' variable must be a string or a list of strings") command += '| tail -n +2' # -header, do not use 'h' return command
[docs] def _get_submit_script_header(self, job_tmpl): """ Return the submit script header, using the parameters from the job_tmpl. Args: job_tmpl: an JobTemplate instance with relevant parameters set. """ # pylint: disable=too-many-branches lines = [] empty_line = '' # Redirecting script output on the correct files # Should be one of the first commands if job_tmpl.sched_output_path: lines.append(f'exec > {job_tmpl.sched_output_path}') if job_tmpl.sched_join_files: # TODO: manual says: # pylint: disable=fixme # By default both standard output and standard error are directed # to a file of the name "slurm-%j.out", where the "%j" is replaced # with the job allocation number. # See that this automatic redirection works also if # I specify a different --output file if job_tmpl.sched_error_path: self.logger.info('sched_join_files is True, but sched_error_path is set; ignoring sched_error_path') else: if job_tmpl.sched_error_path: lines.append(f'exec 2> {job_tmpl.sched_error_path}') else: # To avoid automatic join of files lines.append('exec 2>&1') if job_tmpl.max_memory_kb: self.logger.warning('Physical memory limiting is not supported by the direct scheduler.') if not job_tmpl.import_sys_environment: lines.append('env --ignore-environment \\') if job_tmpl.custom_scheduler_commands: lines.append(job_tmpl.custom_scheduler_commands) if job_tmpl.job_resource and job_tmpl.job_resource.num_cores_per_mpiproc: lines.append(f'export OMP_NUM_THREADS={job_tmpl.job_resource.num_cores_per_mpiproc}') if job_tmpl.job_environment: lines.append(self._get_submit_script_environment_variables(job_tmpl)) if job_tmpl.rerunnable: self.logger.warning( "The 'rerunnable' option is set to 'True', but has no effect when using the direct scheduler." ) lines.append(empty_line) ## The following code is not working as there's an empty line ## inserted between the header and the actual command. # if job_tmpl.max_wallclock_seconds is not None: # try: # tot_secs = int(job_tmpl.max_wallclock_seconds) # if tot_secs <= 0: # raise ValueError # except ValueError: # raise ValueError( # "max_wallclock_seconds must be " # "a positive integer (in seconds)! It is instead '{}'" # "".format((job_tmpl.max_wallclock_seconds))) # lines.append("timeout {} \\".format(tot_secs)) return '\n'.join(lines)
[docs] def _get_submit_command(self, submit_script): """ Return the string to execute to submit a given script. .. note:: One needs to redirect stdout and stderr to /dev/null otherwise the daemon remains hanging for the script to run :param submit_script: the path of the submit script relative to the working directory. IMPORTANT: submit_script should be already escaped. """ submit_command = f'bash {submit_script} > /dev/null 2>&1 & echo $!' self.logger.info(f'submitting with: {submit_command}') return submit_command
[docs] def _parse_joblist_output(self, retval, stdout, stderr): """ Parse the queue output string, as returned by executing the command returned by _get_joblist_command command (qstat -f). Return a list of JobInfo objects, one of each job, each relevant parameters implemented. .. note:: depending on the scheduler configuration, finished jobs may either appear here, or not. This function will only return one element for each job find in the qstat output; missing jobs (for whatever reason) simply will not appear here. """ import re filtered_stderr = '\n'.join(l for l in stderr.split('\n')) if filtered_stderr.strip(): self.logger.warning(f"Warning in _parse_joblist_output, non-empty (filtered) stderr='{filtered_stderr}'") if retval != 0: raise SchedulerError('Error during direct execution parsing (_parse_joblist_output function)') # Create dictionary and parse specific fields job_list = [] for line in stdout.split('\n'): if re.search(r'^\s*PID', line) or line == '': # Skip the header if present continue line = re.sub(r'^\s+', '', line) job = re.split(r'\s+', line) this_job = JobInfo() this_job.job_id = job[0] if len(job) < 3: raise SchedulerError(f"Unexpected output from the scheduler, not enough fields in line '{line}'") try: job_state_string = job[1][0] # I just check the first character except IndexError: self.logger.debug(f"No 'job_state' field for job id {this_job.job_id}") this_job.job_state = JobState.UNDETERMINED else: try: this_job.job_state = \ _MAP_STATUS_PS[job_state_string] except KeyError: self.logger.warning(f"Unrecognized job_state '{job_state_string}' for job id {this_job.job_id}") this_job.job_state = JobState.UNDETERMINED try: # I strip the part after the @: is this always ok? this_job.job_owner = job[2] except KeyError: self.logger.debug(f"No 'job_owner' field for job id {this_job.job_id}") try: this_job.wallclock_time_seconds = self._convert_time(job[3]) except KeyError: # May not have started yet pass except ValueError: self.logger.warning(f"Error parsing 'resources_used.walltime' for job id {this_job.job_id}") # I append to the list of jobs to return job_list.append(this_job) return job_list
[docs] def get_jobs(self, jobs=None, user=None, as_dict=False): """ Overrides original method from DirectScheduler in order to list missing processes as DONE. """ job_stats = super().get_jobs(jobs=jobs, user=user, as_dict=as_dict) found_jobs = [] # Get the list of known jobs if as_dict: found_jobs = job_stats.keys() else: found_jobs = [j.job_id for j in job_stats] # Now check if there are any the user requested but were not found not_found_jobs = list(set(jobs) - set(found_jobs)) if jobs else [] for job_id in not_found_jobs: job = JobInfo() job.job_id = job_id job.job_state = JobState.DONE # Owner and wallclock time is unknown if as_dict: job_stats[job_id] = job else: job_stats.append(job) return job_stats
[docs] def _convert_time(self, string): """ Convert a string in the format HH:MM:SS to a number of seconds. """ import re pieces = re.split('[:.]', string) if len(pieces) != 3: self.logger.warning(f'Wrong number of pieces (expected 3) for time string {string}') raise ValueError('Wrong number of pieces for time string.') days = 0 pieces_first = pieces[0].split('-') if len(pieces_first) == 2: days, pieces[0] = pieces_first days = int(days) try: hours = int(pieces[0]) if hours < 0: raise ValueError except ValueError: self.logger.warning(f'Not a valid number of hours: {pieces[0]}') raise ValueError('Not a valid number of hours.') try: mins = int(pieces[1]) if mins < 0: raise ValueError except ValueError: self.logger.warning(f'Not a valid number of minutes: {pieces[1]}') raise ValueError('Not a valid number of minutes.') try: secs = int(pieces[2]) if secs < 0: raise ValueError except ValueError: self.logger.warning(f'Not a valid number of seconds: {pieces[2]}') raise ValueError('Not a valid number of seconds.') return days * 86400 + hours * 3600 + mins * 60 + secs
[docs] def _parse_submit_output(self, retval, stdout, stderr): """ Parse the output of the submit command, as returned by executing the command returned by _get_submit_command command. To be implemented by the plugin. Return a string with the JobID. """ if retval != 0: self.logger.error(f'Error in _parse_submit_output: retval={retval}; stdout={stdout}; stderr={stderr}') raise SchedulerError(f'Error during submission, retval={retval}\nstdout={stdout}\nstderr={stderr}') if stderr.strip(): self.logger.warning( f'in _parse_submit_output for {str(self.transport)}: there was some text in stderr: {stderr}' ) if not stdout.strip(): self.logger.debug(f'Unable to get the PID: retval={retval}; stdout={stdout}; stderr={stderr}') raise SchedulerError(f'Unable to get the PID: retval={retval}; stdout={stdout}; stderr={stderr}') return stdout.strip()
[docs] def _get_kill_command(self, jobid): """ Return the command to kill the job with specified jobid. """ submit_command = f'kill {jobid}' self.logger.info(f'killing job {jobid}') return submit_command
[docs] def _parse_kill_output(self, retval, stdout, stderr): """ Parse the output of the kill command. To be implemented by the plugin. :return: True if everything seems ok, False otherwise. """ if retval != 0: self.logger.error(f'Error in _parse_kill_output: retval={retval}; stdout={stdout}; stderr={stderr}') return False if stderr.strip(): self.logger.warning( f'in _parse_kill_output for {str(self.transport)}: there was some text in stderr: {stderr}' ) if stdout.strip(): self.logger.warning( f'in _parse_kill_output for {str(self.transport)}: there was some text in stdout: {stdout}' ) return True