def recover(self, job, job_wrapper):
"""Recovers jobs stuck in the queued/running state when Galaxy started"""
job_id = job.get_job_runner_external_id()
pbs_job_state = AsynchronousJobState()
pbs_job_state.output_file = "%s/%s.o" % (
self.app.config.cluster_files_directory, job.id)
pbs_job_state.error_file = "%s/%s.e" % (
self.app.config.cluster_files_directory, job.id)
pbs_job_state.exit_code_file = "%s/%s.ec" % (
self.app.config.cluster_files_directory, job.id)
pbs_job_state.job_file = "%s/%s.sh" % (
self.app.config.cluster_files_directory, job.id)
pbs_job_state.job_id = str(job_id)
pbs_job_state.runner_url = job_wrapper.get_job_runner_url()
pbs_job_state.job_destination = job_wrapper.job_destination
job_wrapper.command_line = job.command_line
pbs_job_state.job_wrapper = job_wrapper
if job.state == model.Job.states.RUNNING:
log.debug(
"(%s/%s) is still in running state, adding to the PBS queue" %
(job.id, job.get_job_runner_external_id()))
pbs_job_state.old_state = 'R'
pbs_job_state.running = True
self.monitor_queue.put(pbs_job_state)
elif job.state == model.Job.states.QUEUED:
log.debug(
"(%s/%s) is still in PBS queued state, adding to the PBS queue"
% (job.id, job.get_job_runner_external_id()))
pbs_job_state.old_state = 'Q'
pbs_job_state.running = False
self.monitor_queue.put(pbs_job_state)
def recover(self, job, job_wrapper):
"""Recovers jobs stuck in the queued/running state when Galaxy started"""
job_id = job.get_job_runner_external_id()
pbs_job_state = AsynchronousJobState()
pbs_job_state.output_file = "%s/%s.o" % (self.app.config.cluster_files_directory, job.id)
pbs_job_state.error_file = "%s/%s.e" % (self.app.config.cluster_files_directory, job.id)
pbs_job_state.exit_code_file = "%s/%s.ec" % (self.app.config.cluster_files_directory, job.id)
pbs_job_state.job_file = "%s/%s.sh" % (self.app.config.cluster_files_directory, job.id)
pbs_job_state.job_id = str(job_id)
pbs_job_state.runner_url = job_wrapper.get_job_runner_url()
pbs_job_state.job_destination = job_wrapper.job_destination
job_wrapper.command_line = job.command_line
pbs_job_state.job_wrapper = job_wrapper
if job.state == model.Job.states.RUNNING:
log.debug(
"(%s/%s) is still in running state, adding to the PBS queue"
% (job.id, job.get_job_runner_external_id())
)
pbs_job_state.old_state = "R"
pbs_job_state.running = True
self.monitor_queue.put(pbs_job_state)
elif job.state == model.Job.states.QUEUED:
log.debug(
"(%s/%s) is still in PBS queued state, adding to the PBS queue"
% (job.id, job.get_job_runner_external_id())
)
pbs_job_state.old_state = "Q"
pbs_job_state.running = False
self.monitor_queue.put(pbs_job_state)
def queue_job(self, job_wrapper):
"""Create PBS script for a job and submit it to the PBS queue"""
# prepare the job
if not self.prepare_job(
job_wrapper,
include_metadata=not (self.app.config.pbs_stage_path)):
return
job_destination = job_wrapper.job_destination
# Determine the job's PBS destination (server/queue) and options from the job destination definition
pbs_queue_name = None
pbs_server_name = self.default_pbs_server
pbs_options = []
if '-q' in job_destination.params and 'destination' not in job_destination.params:
job_destination.params['destination'] = job_destination.params.pop(
'-q')
if 'destination' in job_destination.params:
if '@' in job_destination.params['destination']:
# Destination includes a server
pbs_queue_name, pbs_server_name = job_destination.params[
'destination'].split('@')
if pbs_queue_name == '':
# e.g. `qsub -q @server`
pbs_queue_name = None
else:
# Destination is just a queue
pbs_queue_name = job_destination.params['destination']
job_destination.params.pop('destination')
# Parse PBS params
pbs_options = self.parse_destination_params(job_destination.params)
# Explicitly set the determined PBS destination in the persisted job destination for recovery
job_destination.params['destination'] = '%s@%s' % (pbs_queue_name or
'', pbs_server_name)
c = pbs.pbs_connect(util.smart_str(pbs_server_name))
if c <= 0:
errno, text = pbs.error()
job_wrapper.fail(
"Unable to queue job for execution. Resubmitting the job may succeed."
)
log.error("Connection to PBS server for submit failed: %s: %s" %
(errno, text))
return
# define job attributes
ofile = "%s/%s.o" % (self.app.config.cluster_files_directory,
job_wrapper.job_id)
efile = "%s/%s.e" % (self.app.config.cluster_files_directory,
job_wrapper.job_id)
ecfile = "%s/%s.ec" % (self.app.config.cluster_files_directory,
job_wrapper.job_id)
output_fnames = job_wrapper.get_output_fnames()
# If an application server is set, we're staging
if self.app.config.pbs_application_server:
pbs_ofile = self.app.config.pbs_application_server + ':' + ofile
pbs_efile = self.app.config.pbs_application_server + ':' + efile
output_files = [str(o) for o in output_fnames]
output_files.append(ecfile)
stagein = self.get_stage_in_out(job_wrapper.get_input_fnames() +
output_files,
symlink=True)
stageout = self.get_stage_in_out(output_files)
attrs = [
dict(name=pbs.ATTR_o, value=pbs_ofile),
dict(name=pbs.ATTR_e, value=pbs_efile),
dict(name=pbs.ATTR_stagein, value=stagein),
dict(name=pbs.ATTR_stageout, value=stageout),
]
# If not, we're using NFS
else:
attrs = [
dict(name=pbs.ATTR_o, value=ofile),
dict(name=pbs.ATTR_e, value=efile),
]
# define PBS job options
attrs.append(
dict(name=pbs.ATTR_N,
value=str("%s_%s_%s" %
(job_wrapper.job_id, job_wrapper.tool.id,
job_wrapper.user))))
job_attrs = pbs.new_attropl(len(attrs) + len(pbs_options))
for i, attr in enumerate(attrs + pbs_options):
job_attrs[i].name = attr['name']
job_attrs[i].value = attr['value']
if 'resource' in attr:
job_attrs[i].resource = attr['resource']
exec_dir = os.path.abspath(job_wrapper.working_directory)
# write the job script
if self.app.config.pbs_stage_path != '':
# touch the ecfile so that it gets staged
with file(ecfile, 'a'):
os.utime(ecfile, None)
stage_commands = pbs_symlink_template % (
" ".join(job_wrapper.get_input_fnames() + output_files),
self.app.config.pbs_stage_path,
exec_dir,
)
else:
stage_commands = ''
env_setup_commands = [stage_commands]
script = self.get_job_file(job_wrapper,
exit_code_path=ecfile,
env_setup_commands=env_setup_commands)
job_file = "%s/%s.sh" % (self.app.config.cluster_files_directory,
job_wrapper.job_id)
self.write_executable_script(job_file, script)
# job was deleted while we were preparing it
if job_wrapper.get_state() == model.Job.states.DELETED:
log.debug(
"Job %s deleted by user before it entered the PBS queue" %
job_wrapper.job_id)
pbs.pbs_disconnect(c)
if self.app.config.cleanup_job in ("always", "onsuccess"):
self.cleanup((ofile, efile, ecfile, job_file))
job_wrapper.cleanup()
return
# submit
# The job tag includes the job and the task identifier
# (if a TaskWrapper was passed in):
galaxy_job_id = job_wrapper.get_id_tag()
log.debug("(%s) submitting file %s" % (galaxy_job_id, job_file))
tries = 0
while tries < 5:
job_id = pbs.pbs_submit(c, job_attrs, job_file, pbs_queue_name,
None)
tries += 1
if job_id:
pbs.pbs_disconnect(c)
break
errno, text = pbs.error()
log.warning("(%s) pbs_submit failed (try %d/5), PBS error %d: %s" %
(galaxy_job_id, tries, errno, text))
time.sleep(2)
else:
log.error("(%s) All attempts to submit job failed" % galaxy_job_id)
job_wrapper.fail(
"Unable to run this job due to a cluster error, please retry it later"
)
return
if pbs_queue_name is None:
log.debug("(%s) queued in default queue as %s" %
(galaxy_job_id, job_id))
else:
log.debug("(%s) queued in %s queue as %s" %
(galaxy_job_id, pbs_queue_name, job_id))
# persist destination
job_wrapper.set_job_destination(job_destination, job_id)
# Store PBS related state information for job
job_state = AsynchronousJobState()
job_state.job_wrapper = job_wrapper
job_state.job_id = job_id
job_state.job_file = job_file
job_state.output_file = ofile
job_state.error_file = efile
job_state.exit_code_file = ecfile
job_state.old_state = 'N'
job_state.running = False
job_state.job_destination = job_destination
# Add to our 'queue' of jobs to monitor
self.monitor_queue.put(job_state)
def queue_job( self, job_wrapper ):
"""Create PBS script for a job and submit it to the PBS queue"""
# prepare the job
if not self.prepare_job( job_wrapper, include_metadata=not( self.app.config.pbs_stage_path ) ):
return
job_destination = job_wrapper.job_destination
# Determine the job's PBS destination (server/queue) and options from the job destination definition
pbs_queue_name = None
pbs_server_name = self.default_pbs_server
pbs_options = []
if '-q' in job_destination.params and 'destination' not in job_destination.params:
job_destination.params['destination'] = job_destination.params.pop('-q')
if 'destination' in job_destination.params:
if '@' in job_destination.params['destination']:
# Destination includes a server
pbs_queue_name, pbs_server_name = job_destination.params['destination'].split('@')
if pbs_queue_name == '':
# e.g. `qsub -q @server`
pbs_queue_name = None
else:
# Destination is just a queue
pbs_queue_name = job_destination.params['destination']
job_destination.params.pop('destination')
# Parse PBS params
pbs_options = self.parse_destination_params(job_destination.params)
# Explicitly set the determined PBS destination in the persisted job destination for recovery
job_destination.params['destination'] = '%s@%s' % (pbs_queue_name or '', pbs_server_name)
c = pbs.pbs_connect( util.smart_str( pbs_server_name ) )
if c <= 0:
errno, text = pbs.error()
job_wrapper.fail( "Unable to queue job for execution. Resubmitting the job may succeed." )
log.error( "Connection to PBS server for submit failed: %s: %s" % ( errno, text ) )
return
# define job attributes
ofile = "%s/%s.o" % (self.app.config.cluster_files_directory, job_wrapper.job_id)
efile = "%s/%s.e" % (self.app.config.cluster_files_directory, job_wrapper.job_id)
ecfile = "%s/%s.ec" % (self.app.config.cluster_files_directory, job_wrapper.job_id)
output_fnames = job_wrapper.get_output_fnames()
# If an application server is set, we're staging
if self.app.config.pbs_application_server:
pbs_ofile = self.app.config.pbs_application_server + ':' + ofile
pbs_efile = self.app.config.pbs_application_server + ':' + efile
output_files = [ str( o ) for o in output_fnames ]
output_files.append(ecfile)
stagein = self.get_stage_in_out( job_wrapper.get_input_fnames() + output_files, symlink=True )
stageout = self.get_stage_in_out( output_files )
attrs = [
dict( name=pbs.ATTR_o, value=pbs_ofile ),
dict( name=pbs.ATTR_e, value=pbs_efile ),
dict( name=pbs.ATTR_stagein, value=stagein ),
dict( name=pbs.ATTR_stageout, value=stageout ),
]
# If not, we're using NFS
else:
attrs = [
dict( name=pbs.ATTR_o, value=ofile ),
dict( name=pbs.ATTR_e, value=efile ),
]
# define PBS job options
attrs.append( dict( name=pbs.ATTR_N, value=str( "%s_%s_%s" % ( job_wrapper.job_id, job_wrapper.tool.id, job_wrapper.user ) ) ) )
job_attrs = pbs.new_attropl( len( attrs ) + len( pbs_options ) )
for i, attr in enumerate( attrs + pbs_options ):
job_attrs[i].name = attr['name']
job_attrs[i].value = attr['value']
if 'resource' in attr:
job_attrs[i].resource = attr['resource']
exec_dir = os.path.abspath( job_wrapper.working_directory )
# write the job script
if self.app.config.pbs_stage_path != '':
# touch the ecfile so that it gets staged
with open(ecfile, 'a'):
os.utime(ecfile, None)
stage_commands = pbs_symlink_template % (
" ".join( job_wrapper.get_input_fnames() + output_files ),
self.app.config.pbs_stage_path,
exec_dir,
)
else:
stage_commands = ''
env_setup_commands = [ stage_commands ]
script = self.get_job_file(job_wrapper, exit_code_path=ecfile, env_setup_commands=env_setup_commands)
job_file = "%s/%s.sh" % (self.app.config.cluster_files_directory, job_wrapper.job_id)
self.write_executable_script( job_file, script )
# job was deleted while we were preparing it
if job_wrapper.get_state() == model.Job.states.DELETED:
log.debug( "Job %s deleted by user before it entered the PBS queue" % job_wrapper.job_id )
pbs.pbs_disconnect(c)
if job_wrapper.cleanup_job in ( "always", "onsuccess" ):
self.cleanup( ( ofile, efile, ecfile, job_file ) )
job_wrapper.cleanup()
return
# submit
# The job tag includes the job and the task identifier
# (if a TaskWrapper was passed in):
galaxy_job_id = job_wrapper.get_id_tag()
log.debug("(%s) submitting file %s" % ( galaxy_job_id, job_file ) )
tries = 0
while tries < 5:
job_id = pbs.pbs_submit(c, job_attrs, job_file, pbs_queue_name, None)
tries += 1
if job_id:
pbs.pbs_disconnect(c)
break
errno, text = pbs.error()
log.warning( "(%s) pbs_submit failed (try %d/5), PBS error %d: %s" % (galaxy_job_id, tries, errno, text) )
time.sleep(2)
else:
log.error( "(%s) All attempts to submit job failed" % galaxy_job_id )
job_wrapper.fail( "Unable to run this job due to a cluster error, please retry it later" )
return
if pbs_queue_name is None:
log.debug("(%s) queued in default queue as %s" % (galaxy_job_id, job_id) )
else:
log.debug("(%s) queued in %s queue as %s" % (galaxy_job_id, pbs_queue_name, job_id) )
# persist destination
job_wrapper.set_job_destination( job_destination, job_id )
# Store PBS related state information for job
job_state = AsynchronousJobState()
job_state.job_wrapper = job_wrapper
job_state.job_id = job_id
job_state.job_file = job_file
job_state.output_file = ofile
job_state.error_file = efile
job_state.exit_code_file = ecfile
job_state.old_state = 'N'
job_state.running = False
job_state.job_destination = job_destination
# Add to our 'queue' of jobs to monitor
self.monitor_queue.put( job_state )
