def main():
config['LOG_FILE'] = '/tmp/oar_kamelot.log'
logger = get_logger("oar.kamelot_fifo", forward_stderr=True)
config.setdefault_config(DEFAULT_CONFIG)
plt = Platform()
schedule_fifo_cycle(plt, "default")
logger.info("That's all folks")
def cli(job, full, state, user, array, compact, gantt, events, properties, accounting, sql, format, json, yaml, version):
config.setdefault_config(DEFAULT_CONFIG)
oarapi = OarApi()
if not job:
answer = oarapi.get('jobs/details.json')
print_jobs(True, answer)
def send_to_hulot(cmd, data):
config.setdefault_config({"FIFO_HULOT": "/tmp/oar_hulot_pipe"})
fifoname = config["FIFO_HULOT"]
try:
with open(fifoname, 'w') as fifo:
fifo.write('HALT:%s\n' % data)
fifo.flush()
except IOError as e:
e.strerror = 'Unable to communication with Hulot: %s (%s)' % fifoname % e.strerror
logger.error(e.strerror)
return 1
return 0
def create_app(**kwargs):
"""Return the OAR API application instance."""
app = Flask(__name__)
app.wsgi_app = WSGIProxyFix(app.wsgi_app)
config.setdefault_config(default_config)
app.config.update(config)
db.query_class = APIQuery
db.query_collection_class = APIQueryCollection
register_error_handlers(app)
register_hooks(app)
register_extensions(app)
register_blueprints(app)
register_proxy(app, **kwargs)
return app
def main():
config['LOG_FILE'] = '/tmp/oar_kamelot.log'
logger = get_logger("oar.kamelot", forward_stderr=True)
config.setdefault_config(DEFAULT_CONFIG)
plt = Platform()
logger.debug("argv..." + str(sys.argv))
if len(sys.argv) > 2:
schedule_cycle(plt, int(float(sys.argv[2])), sys.argv[1])
elif len(sys.argv) == 2:
schedule_cycle(plt, plt.get_time(), sys.argv[1])
else:
schedule_cycle(plt, plt.get_time())
logger.info("That's all folks")
from oar.lib import db
db.commit()
from pwd import getpwnam
import pdb
# Set undefined config value to default one
DEFAULT_CONFIG = {
'META_SCHED_CMD': 'kao',
'SERVER_HOSTNAME': 'localhost',
'APPENDICE_SERVER_PORT': '6668', #new endpoint which replace appendic
'SCHEDULER_MIN_TIME_BETWEEN_2_CALLS': '1',
'FINAUD_FREQUENCY': '300',
'LOG_FILE': '/var/log/oar.log',
'ENERGY_SAVING_INTERNAL': 'no'
}
config.setdefault_config(DEFAULT_CONFIG)
# retrieve umask and set new one
old_umask = os.umask(0o022)
# TODO
# my $oldfh = select(STDERR); $| = 1; select($oldfh);
# $oldfh = select(STDOUT); $| = 1; select($oldfh);
# Everything is run by oar user (The real uid of this process.)
#os.environ['OARDO_UID'] = str(os.geteuid())
# TODO
# my $Redirect_STD_process = OAR::Modules::Judas::redirect_everything();
from copy import deepcopy
import oar.kao.scheduling
from oar.lib import config
from oar.lib.interval import (intersec, itvs_size, extract_n_scattered_block_itv,
aggregate_itvs)
import pickle
try:
import zerorpc
except ImportError:
zerorpc = None
# Set undefined config value to default one
config.setdefault_config({
'COORM_DEFAULT_TIMEOUT': 10,
})
def find_default(itvs_avail, hy_res_rqts, hy, beginning, *find_args, **find_kwargs):
"""Simple wrap function to default function for test purpose"""
return oar.kao.scheduling.find_resource_hierarchies_job(itvs_avail, hy_res_rqts, hy)
def assign_default(slots_set, job, hy, min_start_time, *assign_args, **assign_kwargs):
"""Simple wrap function to default function for test purpose"""
return oar.kao.scheduling.assign_resources_mld_job_split_slots(slots_set, job, hy, min_start_time)
def find_begin(itvs_avail, hy_res_rqts, hy, beginning, *find_args, **find_kwargs):
"""Simple function to test beginning value which is set to True is the slot begins the slotset (slotset.begin == slots[1].b).
def test_simsim_1():
# Set undefined config value to default one
DEFAULT_CONFIG = {
'HIERARCHY_LABELS': 'resource_id,network_address',
'SCHEDULER_RESOURCE_ORDER': "resource_id ASC",
'SCHEDULER_JOB_SECURITY_TIME': '60',
'SCHEDULER_AVAILABLE_SUSPENDED_RESOURCE_TYPE': 'default',
'FAIRSHARING_ENABLED': 'no',
'SCHEDULER_FAIRSHARING_MAX_JOB_PER_USER': '******',
'QUOTAS': 'no'
}
config.setdefault_config(DEFAULT_CONFIG)
nb_res = 32
#
# generate ResourceSet
#
hy_resource_id = [[(i, i)] for i in range(1, nb_res + 1)]
res_set = ResourceSetSimu(
rid_i2o=range(nb_res + 1),
rid_o2i=range(nb_res + 1),
roid_itvs=[(1, nb_res)],
hierarchy={'resource_id': hy_resource_id},
available_upto={2147483600: [(1, nb_res)]}
)
#
# generate jobs
#
nb_jobs = 4
jobs = {}
submission_time_jids = []
for i in range(1, nb_jobs + 1):
jobs[i] = JobSimu(id=i,
state="Waiting",
queue="test",
start_time=0,
walltime=0,
types={},
res_set=[],
moldable_id=0,
mld_res_rqts=[
(i, 60, [([("resource_id", 15)], [(0, nb_res - 1)])])],
run_time=20 * i,
deps=[],
key_cache={},
ts=False, ph=0,
assign=False, find=False)
submission_time_jids.append((10, [i]))
# submission_time_jids= [(10, [1,2,3,4])]
# submission_time_jids= [(10, [1,2]), (10, [3])]
print(submission_time_jids)
simsched = SimSched(res_set, jobs, submission_time_jids)
simsched.run()
plt = simsched.platform
print("Number completed jobs:", len(plt.completed_jids))
print("Completed job ids:", plt.completed_jids)
print(jobs)
# assert True
assert len(plt.completed_jids) == nb_jobs
def karma_jobs_sorting(queue, now, jids, jobs, plt):
# if "SCHEDULER_FAIRSHARING_MAX_JOB_PER_USER" in config:
# fairsharing_nb_job_limit = config["SCHEDULER_FAIRSHARING_MAX_JOB_PER_USER"]
# TODO NOT UDSED
# fairsharing_nb_job_limit = 100000
karma_window_size = 3600 * 30 * 24 # TODO in conf ???
# Set undefined config value to default one
default_config = {
"SCHEDULER_FAIRSHARING_PROJECT_TARGETS": "{default => 21.0}",
"SCHEDULER_FAIRSHARING_USER_TARGETS": "{default => 22.0}",
"SCHEDULER_FAIRSHARING_COEF_PROJECT": "0",
"SCHEDULER_FAIRSHARING_COEF_USER": "******",
"SCHEDULER_FAIRSHARING_COEF_USER_ASK": "1"
}
config.setdefault_config(default_config)
# get fairsharing config if any
karma_proj_targets = perl_hash_2_dict(
config["SCHEDULER_FAIRSHARING_PROJECT_TARGETS"])
karma_user_targets = perl_hash_2_dict(
config["SCHEDULER_FAIRSHARING_USER_TARGETS"])
karma_coeff_proj_consumption = float(
config["SCHEDULER_FAIRSHARING_COEF_PROJECT"])
karma_coeff_user_consumption = float(
config["SCHEDULER_FAIRSHARING_COEF_USER"])
karma_coeff_user_asked_consumption = float(
config["SCHEDULER_FAIRSHARING_COEF_USER_ASK"])
#
# Sort jobs accordingly to karma value (fairsharing) *)
# *)
window_start = now - karma_window_size
window_stop = now
karma_sum_time_asked, karma_sum_time_used = plt.get_sum_accounting_window(
queue, window_start, window_stop)
karma_projects_asked, karma_projects_used = plt.get_sum_accounting_by_project(
queue, window_start, window_stop)
karma_users_asked, karma_users_used = plt.get_sum_accounting_by_user(
queue, window_start, window_stop)
#
# compute karma for each job
#
for job in itervalues(jobs):
if job.project in karma_projects_used:
karma_proj_used_j = karma_projects_used[job.project]
else:
karma_proj_used_j = 0.0
if job.user in karma_users_used:
karma_user_used_j = karma_users_used[job.user]
else:
karma_user_used_j = 0.0
if job.user in karma_users_asked:
karma_user_asked_j = karma_users_asked[job.user]
else:
karma_user_asked_j = 0.0
if job.project in karma_proj_targets:
karma_proj_target = karma_proj_targets[job.project]
else:
karma_proj_target = 0.0
if job.user in karma_user_targets:
karma_user_target = karma_user_targets[job.user] / 100.0
else:
karma_user_target = 0.0
# x1 = karma_coeff_proj_consumption * ((karma_proj_used_j / karma_sum_time_used) - (karma_proj_target / 100.0))
# x2 = karma_coeff_user_consumption * ((karma_user_used_j / karma_sum_time_used) - (karma_user_target / 100.0))
# x3 = karma_coeff_user_asked_consumption * ((karma_user_asked_j / karma_sum_time_asked) - (karma_user_target / 100.0))
# print "yopypop", x1, x2, x3
job.karma = (karma_coeff_proj_consumption * ((karma_proj_used_j / karma_sum_time_used) - (karma_proj_target / 100.0)) +
karma_coeff_user_consumption * ((karma_user_used_j / karma_sum_time_used) - (karma_user_target / 100.0)) +
karma_coeff_user_asked_consumption *
((karma_user_asked_j / karma_sum_time_asked) -
(karma_user_target / 100.0))
)
# print "job.karma", job.karma
# sort jids according to jobs' karma value
# print jids
karma_ordered_jids = sorted(jids, key=lambda jid: jobs[jid].karma)
# print karma_ordered_jids
return karma_ordered_jids
def main():
logger = get_logger("oar.kamelot_basic", forward_stderr=True)
config.setdefault_config(DEFAULT_CONFIG)
plt = Platform()
schedule_cycle(plt)
logger.info("That's all folks")
def cli(
command,
interactive,
queue,
resource,
reservation,
connect,
type,
checkpoint,
property,
resubmit,
scanscript,
project,
signal,
directory,
name,
after,
notify,
array,
array_param_file,
use_job_key,
import_job_key_from_file,
import_job_key_inline,
export_job_key_to_file,
stdout,
stderr,
hold,
version,
):
"""Submit a job to OAR batch scheduler."""
config.setdefault_config(DEFAULT_CONFIG)
# import pdb; pdb.set_trace()
# print(resource)
# When the walltime of a job is not defined
default_job_walltime = str(config["DEFAULT_JOB_WALLTIME"])
log_warning = "" # TODO
log_error = ""
log_info = ""
log_std = ""
remote_host = config["SERVER_HOSTNAME"]
remote_port = int(config["SERVER_PORT"])
if "OARSUB_DEFAULT_RESOURCES" in config:
default_resources = config["OARSUB_DEFAULT_RESOURCES"]
else:
default_resources = "/resource_id=1"
if "OARSUB_NODES_RESOURCES" in config:
nodes_resources = config["OARSUB_NODES_RESOURCES"]
else:
nodes_resources = "resource_id"
# TODO Deploy_hostname / Cosystem_hostname
# $Deploy_hostname = get_conf("DEPLOY_HOSTNAME");
# if (!defined($Deploy_hostname)){
# $Deploy_hostname = $remote_host;
# }
# $Cosystem_hostname = get_conf("COSYSTEM_HOSTNAME");
# if (!defined($Cosystem_hostname)){
# $Cosystem_hostname = $remote_host;
# }
cpuset_field = config["JOB_RESOURCE_MANAGER_PROPERTY_DB_FIELD"]
cpuset_path = config["CPUSET_PATH"]
if "OAR_RUNTIME_DIRECTORY" in config:
pass
# if (is_conf("OAR_RUNTIME_DIRECTORY")){
# OAR::Sub::set_default_oarexec_directory(get_conf("OAR_RUNTIME_DIRECTORY"));
# }
# my $default_oar_dir = OAR::Sub::get_default_oarexec_directory();
# if (!(((-d $default_oar_dir) and (-O $default_oar_dir)) or (mkdir($default_oar_dir)))){
# die("# Error: failed to create the OAR directory $default_oar_dir, or bad permissions.\n");
# }
binpath = ""
if "OARDIR" in os.environ:
binpath = os.environ["OARDIR"] + "/"
else:
print_error("OARDIR environment variable is not defined.")
sub_exit(1)
openssh_cmd = config["OPENSSH_CMD"]
ssh_timeout = int(config["OAR_SSH_CONNECTION_TIMEOUT"])
# if (is_conf("OAR_SSH_CONNECTION_TIMEOUT")){
# OAR::Sub::set_ssh_timeout(get_conf("OAR_SSH_CONNECTION_TIMEOUT"));
# }
# OAR version
# TODO: OAR is now a set of composition...
#
types = type
properties = lstrip_none(property)
if not directory:
launching_directory = ""
else:
launching_directory = lstrip_none(directory)
initial_request = " ".join(sys.argv[1:])
queue_name = lstrip_none(queue)
reservation_date = lstrip_none(reservation)
if reservation_date:
m = re.search(r"^\s*(\d{4}\-\d{1,2}\-\d{1,2})\s+(\d{1,2}:\d{1,2}:\d{1,2})\s*$", reservation)
if m:
reservation_date = sql_to_local(m.group(1) + " " + m.group(2))
else:
print_error(
'syntax error for the advance reservation start date \
specification. Expected format is:"YYYY-MM-DD hh:mm:ss"'
)
sub_exit(7)
if array:
array_nb = array
else:
array_nb = 1
# Check the default name of the key if we have to generate it
if ("OARSUB_FORCE_JOB_KEY" in config) and config["OARSUB_FORCE_JOB_KEY"] == "yes":
use_job_key = True
else:
use_job_key = False
# TODO ssh_private_key, ssh_public_key,
# ssh_private_key = ''
# ssh_public_key = ''
# TODO import_job_key_file, export_job_key_file
import_job_key_file = ""
export_job_key_file = ""
if resubmit:
print("# Resubmitting job ", resubmit, "...")
ret = resubmit_job(resubmit)
if ret > 0:
job_id = ret
print(" done.\n")
print("OAR_JOB_ID=" + str(job_id))
if signal_almighty(remote_host, remote_port, "Qsub") > 0:
print_error(
"cannot connect to executor "
+ str(remote_host)
+ ":"
+ str(remote_port)
+ ". OAR server might be down."
)
sub_exit(3)
else:
sub_exit(0)
else:
print(" error.")
if ret == -1:
print_error("interactive jobs and advance reservations cannot be resubmitted.")
elif ret == -2:
print_error("only jobs in the Error or Terminated state can be resubmitted.")
elif ret == -3:
print_error("resubmitted job user mismatch.")
elif ret == -4:
print_error("another active job is using the same job key.")
else:
print_error("unknown error.")
sub_exit(4)
if not command and not interactive and not reservation and not connect:
usage()
sub_exit(5)
if interactive and reservation:
print_error("an advance reservation cannot be interactive.")
usage()
sub_exit(7)
if interactive and any(re.match(r"^desktop_computing$", t) for t in type):
print_error(" a desktop computing job cannot be interactive")
usage()
sub_exit(17)
if any(re.match(r"^noop$", t) for t in type):
if interactive:
print_error("a NOOP job cannot be interactive.")
sub_exit(17)
elif connect:
print_error("a NOOP job does not have a shell to connect to.")
sub_exit(17)
# notify : check insecure character
if notify and re.match(r"^.*exec\s*:.+$"):
m = re.search(r".*exec\s*:([a-zA-Z0-9_.\/ -]+)$", notify)
if not m:
print_error(
"insecure characters found in the notification method \
(the allowed regexp is: [a-zA-Z0-9_.\/ -]+)."
)
sub_exit(16)
# TODO Connect to a reservation
# Connect to a reservation
# if (defined($connect_job)){
# Do not kill the job if the user close the window
# $SIG{HUP} = 'DEFAULT';
# OAR::Sub::close_db_connection(); exit(connect_job($connect_job,0,$Openssh_cmd));
# }
if not project:
project = DEFAULT_VALUE["project"]
if not signal:
signal = DEFAULT_VALUE["signal"]
if not directory:
directory = DEFAULT_VALUE["directory"]
resource_request = parse_resource_descriptions(resource, default_resources, nodes_resources)
job_vars = {
"job_type": None,
"resource_request": resource_request,
"command": command,
"info_type": None,
"queue_name": queue_name,
"properties": properties,
"checkpoint": checkpoint,
"signal": signal,
"notify": notify,
"name": name,
"types": types,
"launching_directory": launching_directory,
"dependencies": after,
"stdout": stdout,
"stderr": stderr,
"hold": hold,
"project": project,
"initial_request": initial_request,
"user": os.environ["OARDO_USER"],
"array_id": 0,
"start_time": "0",
"reservation_field": None,
}
if not interactive and command:
cmd_executor = "Qsub"
if scanscript:
# TODO scanscript
pass
array_params = []
if array_param_file:
pass
# TODO
# $array_params_ref = OAR::Sub::read_array_param_file($array_param_file);
# $array_nb = scalar @{$array_params_ref};
if array_nb == 0:
print_error("an array of job must have a number of sub-jobs greater than 0.")
usage()
sub_exit(6)
job_vars["info_type"] = "$Host:$server_port" # TODO "$Host:$server_port"
job_vars["job_type"] = "PASSIVE"
(err, job_id_lst) = add_micheline_jobs(
job_vars,
reservation_date,
use_job_key,
import_job_key_inline,
import_job_key_file,
export_job_key_file,
initial_request,
array_nb,
array_params,
)
else:
# TODO interactive
if command:
print_warning("asking for an interactive job (-I), so ignoring arguments: " + command + " .")
cmd_executor = "Qsub -I"
if array_param_file:
print_error("a array job with parameters given in a file cannot be interactive.")
usage()
sub_exit(9)
if array_nb != 1:
print_error("an array job cannot be interactive.")
usage()
sub_exit(8)
if reservation:
# Test if this job is a reservation and the syntax is right
# TODO Pass
pass
socket_server = init_tcp_server()
(server, server_port) = socket_server.getsockname()
job_vars["info_type"] = server + ":" + str(server_port)
job_vars["job_type"] = "INTERACTIVE"
(err, job_id_lst) = add_micheline_jobs(
job_vars,
reservation_date,
use_job_key,
import_job_key_inline,
import_job_key_file,
export_job_key_file,
initial_request,
array_nb,
array_params,
)
# pdb.set_trace()
if err != 0:
print_error("command failed, please verify your syntax.")
sub_exit(err, "")
oar_array_id = 0
# Print job_id list
if len(job_id_lst) == 1:
print("OAR_JOB_ID=", job_id_lst[0])
else:
job = db["Job"].query.filter(Job.id == job_id_lst[0]).one()
oar_array_id = job.array_id
for job_id in job_id_lst:
print("OAR_JOB_ID=", job_id)
result = (job_id_lst, oar_array_id)
# Notify Almigthy
tools.create_almighty_socket()
tools.notify_almighty(cmd_executor)
if reservation:
# Reservation mode
print_info("advance reservation request: waiting for approval from the scheduler...")
(conn, address) = socket_server.accept()
answer = conn.recv(1024)
if answer[:-1] == "GOOD RESERVATION":
print_info("advance reservation is GRANTED.")
else:
print_info("advance reservation is REJECTED ", answer[:-1])
sub_exit(10)
elif interactive:
# Interactive mode
print_info("interactive mode: waiting...")
prev_str = ""
while True:
(conn, address) = socket_server.accept()
answer = conn.recv(1024)
answer = answer[:-1]
m = re.search(r"\](.*)$", answer)
if m and m.group(1) != prev_str:
print_info(answer)
prev_str = m.group(1)
elif answer != "GOOD JOB":
print_info(answer)
if (
(answer == "GOOD JOB")
or (answer == "BAD JOB")
or (answer == "JOB KILLE")
or re.match(r"^ERROR", answer)
):
break
if answer == "GOOD JOB":
# TODO exit(connect_job($Job_id_list_ref->[0],1,$Openssh_cmd));
pass
else:
sub_exit(11)
sub_exit(0, result)
