def get_slurm_permssions():
try:
shell_string = "sacctmgr show user ${USER} -Pn"
log.debug(shell_string)
lmutil_return = subprocess.check_output(shell_string, shell=True).decode("utf-8").strip().split("|")[-1].lower() # Removed .decode("utf-8") as threw error.
except Exception as details:
log.error("Failed to fetch user permissions, assuming none: " + str(details))
else:
log.info("User SLURM permissions are '" + lmutil_return + "'")
return lmutil_return
def readmake_json(path, default={}):
"""Reads and returns JSON file as dictionary, if none exists one will be created with default value."""
if not os.path.exists(path):
log.error("No file at path '" + path + "'.")
with open(path, "w+") as json_file:
json_file.write(json.dumps(default))
log.error("Empty file created")
with open(path) as json_file:
log.debug(path + " loaded.")
return json.load(json_file)
def ex_slurm_command(sub_input, level="administrator"):
log.debug("Attempting to run SLURM command '" + sub_input + "'.")
if (level == "administrator" and slurm_permissions == "administrator") or (level == "operator" and (slurm_permissions == "operator" or slurm_permissions == "administrator")):
try:
output = subprocess.check_output(sub_input, shell=True).decode("utf-8")
except Exception as details:
log.error("Failed to execute SLURM command '" + sub_input + "':" + str(details))
else:
log.debug("Success!")
time.sleep(5) # Avoid spamming database
return output
else:
with open("run_as_admin.sh", "a+") as f:
f.write(sub_input + "\n")
log.error("User does not have appropriate SLURM permissions to run this command. Writing command to 'run_as_admin.sh'")
def _do_maths(feature):
log.debug("Doing maths...")
hour_index = datetime.datetime.now().hour - 1
feature["history"].append(feature["usage_all"])
while len(feature["history"]) > settings["history_points"]:
feature["history"].pop(0)
# Update average
if feature["hourly_averages"][hour_index]:
feature["hourly_averages"][hour_index] = round(
((feature["usage_all"] * settings["point_weight"]) + (feature["hourly_averages"][hour_index] * (1 - settings["point_weight"]))), 2
)
else:
feature["hourly_averages"][hour_index] = feature["usage_all"]
if not feature["slurm_active"]:
feature["token_soak"] = "--"
log.debug("Skipping " + feature["feature_name"] + " disabled")
else:
feature["token_soak"] = int(min(max(max(feature["history"]), feature["usage_all"]) + feature["buffer_constant"], feature["total"]))
def apply_soak():
def _do_maths(feature):
log.debug("Doing maths...")
hour_index = datetime.datetime.now().hour - 1
feature["history"].append(feature["usage_all"])
while len(feature["history"]) > settings["history_points"]:
feature["history"].pop(0)
# Update average
if feature["hourly_averages"][hour_index]:
feature["hourly_averages"][hour_index] = round(
((feature["usage_all"] * settings["point_weight"]) + (feature["hourly_averages"][hour_index] * (1 - settings["point_weight"]))), 2
)
else:
feature["hourly_averages"][hour_index] = feature["usage_all"]
if not feature["slurm_active"]:
feature["token_soak"] = "--"
log.debug("Skipping " + feature["feature_name"] + " disabled")
else:
feature["token_soak"] = int(min(max(max(feature["history"]), feature["usage_all"]) + feature["buffer_constant"], feature["total"]))
def _update_res(cluster, soak):
log.info("Attempting to update " + cluster + " reservation.")
sub_input = "scontrol update -M " + cluster + " ReservationName=" + res_name + " " + soak
ex_slurm_command(sub_input, "operator")
def _create_res(cluster, soak):
log.info("Attempting to update " + cluster + " reservation.")
sub_input = "scontrol create -M " + cluster + " ReservationName=" + res_name + " StartTime=now Duration=infinite Users=root Flags=LICENSE_ONLY " + soak
ex_slurm_command(sub_input)
if os.environ.get("SOAK", "").lower() == "false":
log.info("Licence Soak skipped due to 'SOAK=FALSE'")
return
log.info("Applying soak...")
res_name = "licence_soak"
res_update_strings = {}
for server in server_list:
for tracked_feature_name, tracked_feature_value in server["tracked_features"].items():
_do_maths(tracked_feature_value)
if not tracked_feature_value["slurm_active"]:
log.debug(tracked_feature_name + "not slurm_active. Skipping...")
continue
if not tracked_feature_value["token_name"]:
log.debug(tracked_feature_name + "has no token name. Skipping...")
continue
for cluster in tracked_feature_value["clusters"]:
if tracked_feature_value["token_soak"]:
if cluster not in res_update_strings:
res_update_strings[cluster] = " licenses="
res_update_strings[cluster] += tracked_feature_value["token_name"] + ":" + str(tracked_feature_value["token_soak"]) + ","
log.debug("Contructing reservation strings")
log.debug(json.dumps(res_update_strings))
for cluster, soak in res_update_strings.items():
if cluster not in settings["clusters"].keys() or "enabled" not in settings["clusters"][cluster].keys() or not settings["clusters"][cluster]["enabled"]:
log.info("Skipping licence soak on " + cluster)
continue
try:
_update_res(cluster, soak)
except Exception as details:
log.error("Reservation update failed: " + str(details))
log.info("Attempting to create new reservation.")
try:
_create_res(cluster, soak)
except Exception as details:
log.error("Failed to create reservation: " + str(details))
else:
log.info("New reservation '" + res_name + "' created successfully.")
else:
log.info(cluster + " reservation updated successfully!")
schedul.enter(settings["squeue_poll_period"], 1, apply_soak)
def ansysli_util():
log.info("Checking Licence Server at '" + server["server"]["address"] + "'... (period " + str(server["server"]["poll_period"]) + "s)")
shell_command_string = "export ANSYSLMD_LICENSE_FILE=$(head -n 1 " + server["licence_file"]["path"] + " | sed -n -e 's/.*=//p');linx64/ansysli_util -liusage"
log.debug(shell_command_string)
sub_return = subprocess.check_output(shell_command_string, shell=True).strip().decode("utf-8", "replace") # Removed .decode("utf-8") as threw error.
log.debug(sub_return)
coarce_pattern = re.compile(r"FEATURENAME: (?P<feature>\S*)\n.*\n.*\n.*\n.*COUNT: (?P<total>\d*)\n.*USED: (?P<count>\d*)", flags=re.M)
fine_pattern = re.compile(r"(?P<user>[A-Za-z0-9]*)@(?P<host>\S*):(?P<pid>\d*)\s*(?P<date>[\d\/]*?) (?P<time>[\d\:]*)\s*(?P<feature>[a-zA-Z_]*)?[^\d]*(?P<count>\d*){1}\s*(?P<version>\S*)", flags=re.M)
#datetime.datetime.now().strftime("%Y-%m-%dT%H:%M:%S")
server["server"]["status"] = "UP" #server_re_match_group["last_stat"]
server["server"]["version"] = ""
# Because ANSYS hates me, ansli_util occasionally records the same licence being used multiple times.
# Track PID list <pid><lic> discard duplicates.
pidpluslic_list=[]
featureanduser_re_match = fine_pattern.finditer(sub_return)
if len(server["tracked_features"].keys()) < 1:
log.warning("No features are being tracked on " + server["server"]["address"])
# Clear previous totals
for tracked_feature in server["tracked_features"].values():
tracked_feature["usage_all"] = 0
tracked_feature["usage_nesi"] = 0
tracked_feature["users_nesi"] = {}
for featureorline in featureanduser_re_match:
group_dic = featureorline.groupdict()
if not group_dic["feature"] or not group_dic["user"]:
continue
# If this is the case, it is a feature header.
match_cluster = cluster_pattern.match(group_dic["host"])
# if duplicate, skip.
if group_dic["pid"]+group_dic["feature"] in pidpluslic_list:
log.debug("Duplicate licence entry.")
continue
else:
pidpluslic_list.append(group_dic["pid"]+group_dic["feature"])
log.debug(json.dumps(group_dic))
if group_dic["feature"] in server["tracked_features"].keys():
log.debug(group_dic["feature"] + " is tracked feature")
# Count gets added regardless of socket
if "count" in group_dic and group_dic["count"].isdigit():
lic_count = int(group_dic["count"])
else:
lic_count = 1
server["tracked_features"][group_dic["feature"]]["usage_all"] += lic_count
#print(server["tracked_features"][group_dic["feature"]]["usage_all"])
# Count gets added regardless of socket
if match_cluster:
#log.debug(group_dic["user"] + " is a user")
# If user not already. Add them.
if group_dic["user"] not in server["tracked_features"][group_dic["feature"]]["users_nesi"]:
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]] = {"count": 0, "tokens": 0, "sockets": [], "soak": 0}
server["tracked_features"][group_dic["feature"]]["usage_nesi"] += lic_count
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]]["count"] += lic_count
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]]["sockets"].append(match_cluster.group(0))
elif match_cluster:
log.info("Untracked feature: " + group_dic["feature"] + " being used by " + group_dic["user"] + " on " + group_dic["host"])
elif group_dic["feature"] in server["untracked_features"]:
log.debug(group_dic["feature"] + " is untracked feature")
else:
server["untracked_features"].append(group_dic["feature"])
log.info("'" + group_dic["feature"] + "' being added to untracked features.")
log.debug(str(pidpluslic_list))
def lmutil():
log.info("Checking Licence Server at '" + server["server"]["address"] + "'... (period " + str(server["server"]["poll_period"]) + "s)")
shell_command_string = poll_methods[server["server"]["poll_method"]]["shell_command"] % server["licence_file"]
log.debug(shell_command_string)
sub_return = subprocess.check_output(shell_command_string, shell=True).strip().decode("utf-8", "replace") # Removed .decode("utf-8") as threw error.
log.debug(sub_return)
# Match server details
server_re_match = poll_methods[server["server"]["poll_method"]]["server_pattern"].search(sub_return)
if server_re_match is None:
return
server_re_match_group = server_re_match.groupdict()
server["server"]["status"] = server_re_match_group["last_stat"]
server["server"]["version"] = server_re_match_group["version"]
log.info("'" + server["server"]["address"] + "' " + server_re_match_group["last_stat"])
# server["server"]["last_time"]=server_re_match_group["last_time"]
featureanduser_re_match = poll_methods[server["server"]["poll_method"]]["licence_pattern"].finditer(sub_return)
if len(server["tracked_features"].keys()) < 1:
log.warning("No features are being tracked on " + server["server"]["address"])
# Clear previous totals
for tracked_feature in server["tracked_features"].values():
tracked_feature["usage_all"] = 0
tracked_feature["usage_nesi"] = 0
tracked_feature["users_nesi"] = {}
last_lic = None
tracked = False
# Read regex by line.
for featureorline in featureanduser_re_match:
group_dic = featureorline.groupdict()
#log.debug(json.dumps(group_dic))
# If this is the case, it is a feature header.
if group_dic["feature"] is not None:
if group_dic["feature"] in server["tracked_features"].keys():
log.debug(group_dic["feature"] + " is tracked feature")
# Last_lic points to the licecne objects.
tracked = True
last_lic = server["tracked_features"][group_dic["feature"]]
if last_lic["total"] != int(group_dic["total"]):
log.warning("total was " + str(last_lic["total"]) + " at last count, now " + str(group_dic["total"]))
last_lic["total"] = int(group_dic["total"])
elif group_dic["feature"] in server["untracked_features"]:
last_lic = group_dic["feature"]
log.debug(group_dic["feature"] + " is untracked feature")
tracked = False
else:
server["untracked_features"].append(group_dic["feature"])
last_lic = group_dic["feature"]
log.info("'" + group_dic["feature"] + "' being added to untracked features.")
tracked = False
continue
# If this is the case, it is a user.
if group_dic["user"] is not None:
#log.debug(group_dic["user"] + " is a user")
match_cluster = cluster_pattern.match(group_dic["host"])
if tracked:
# Count gets added regardless of socket
if "count" in group_dic and group_dic["count"].isdigit():
lic_count = int(group_dic["count"])
else:
lic_count = 1
last_lic["usage_all"] += lic_count
# Count gets added regardless of socket
if match_cluster:
# If user not already. Add them.
if group_dic["user"] not in last_lic["users_nesi"]:
last_lic["users_nesi"][group_dic["user"]] = {"count": 0, "tokens": 0, "sockets": [], "soak": 0}
last_lic["usage_nesi"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["count"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["soak"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["sockets"].append(match_cluster.group(0))
else:
if match_cluster:
log.info("Untracked feature: " + " being used by " + " on ")
def poll_remote(server):
def lmutil():
log.info("Checking Licence Server at '" + server["server"]["address"] + "'... (period " + str(server["server"]["poll_period"]) + "s)")
shell_command_string = poll_methods[server["server"]["poll_method"]]["shell_command"] % server["licence_file"]
log.debug(shell_command_string)
sub_return = subprocess.check_output(shell_command_string, shell=True).strip().decode("utf-8", "replace") # Removed .decode("utf-8") as threw error.
log.debug(sub_return)
# Match server details
server_re_match = poll_methods[server["server"]["poll_method"]]["server_pattern"].search(sub_return)
if server_re_match is None:
return
server_re_match_group = server_re_match.groupdict()
server["server"]["status"] = server_re_match_group["last_stat"]
server["server"]["version"] = server_re_match_group["version"]
log.info("'" + server["server"]["address"] + "' " + server_re_match_group["last_stat"])
# server["server"]["last_time"]=server_re_match_group["last_time"]
featureanduser_re_match = poll_methods[server["server"]["poll_method"]]["licence_pattern"].finditer(sub_return)
if len(server["tracked_features"].keys()) < 1:
log.warning("No features are being tracked on " + server["server"]["address"])
# Clear previous totals
for tracked_feature in server["tracked_features"].values():
tracked_feature["usage_all"] = 0
tracked_feature["usage_nesi"] = 0
tracked_feature["users_nesi"] = {}
last_lic = None
tracked = False
# Read regex by line.
for featureorline in featureanduser_re_match:
group_dic = featureorline.groupdict()
#log.debug(json.dumps(group_dic))
# If this is the case, it is a feature header.
if group_dic["feature"] is not None:
if group_dic["feature"] in server["tracked_features"].keys():
log.debug(group_dic["feature"] + " is tracked feature")
# Last_lic points to the licecne objects.
tracked = True
last_lic = server["tracked_features"][group_dic["feature"]]
if last_lic["total"] != int(group_dic["total"]):
log.warning("total was " + str(last_lic["total"]) + " at last count, now " + str(group_dic["total"]))
last_lic["total"] = int(group_dic["total"])
elif group_dic["feature"] in server["untracked_features"]:
last_lic = group_dic["feature"]
log.debug(group_dic["feature"] + " is untracked feature")
tracked = False
else:
server["untracked_features"].append(group_dic["feature"])
last_lic = group_dic["feature"]
log.info("'" + group_dic["feature"] + "' being added to untracked features.")
tracked = False
continue
# If this is the case, it is a user.
if group_dic["user"] is not None:
#log.debug(group_dic["user"] + " is a user")
match_cluster = cluster_pattern.match(group_dic["host"])
if tracked:
# Count gets added regardless of socket
if "count" in group_dic and group_dic["count"].isdigit():
lic_count = int(group_dic["count"])
else:
lic_count = 1
last_lic["usage_all"] += lic_count
# Count gets added regardless of socket
if match_cluster:
# If user not already. Add them.
if group_dic["user"] not in last_lic["users_nesi"]:
last_lic["users_nesi"][group_dic["user"]] = {"count": 0, "tokens": 0, "sockets": [], "soak": 0}
last_lic["usage_nesi"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["count"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["soak"] += lic_count
last_lic["users_nesi"][group_dic["user"]]["sockets"].append(match_cluster.group(0))
else:
if match_cluster:
log.info("Untracked feature: " + " being used by " + " on ")
def ansysli_util():
log.info("Checking Licence Server at '" + server["server"]["address"] + "'... (period " + str(server["server"]["poll_period"]) + "s)")
shell_command_string = "export ANSYSLMD_LICENSE_FILE=$(head -n 1 " + server["licence_file"]["path"] + " | sed -n -e 's/.*=//p');linx64/ansysli_util -liusage"
log.debug(shell_command_string)
sub_return = subprocess.check_output(shell_command_string, shell=True).strip().decode("utf-8", "replace") # Removed .decode("utf-8") as threw error.
log.debug(sub_return)
coarce_pattern = re.compile(r"FEATURENAME: (?P<feature>\S*)\n.*\n.*\n.*\n.*COUNT: (?P<total>\d*)\n.*USED: (?P<count>\d*)", flags=re.M)
fine_pattern = re.compile(r"(?P<user>[A-Za-z0-9]*)@(?P<host>\S*):(?P<pid>\d*)\s*(?P<date>[\d\/]*?) (?P<time>[\d\:]*)\s*(?P<feature>[a-zA-Z_]*)?[^\d]*(?P<count>\d*){1}\s*(?P<version>\S*)", flags=re.M)
#datetime.datetime.now().strftime("%Y-%m-%dT%H:%M:%S")
server["server"]["status"] = "UP" #server_re_match_group["last_stat"]
server["server"]["version"] = ""
# Because ANSYS hates me, ansli_util occasionally records the same licence being used multiple times.
# Track PID list <pid><lic> discard duplicates.
pidpluslic_list=[]
featureanduser_re_match = fine_pattern.finditer(sub_return)
if len(server["tracked_features"].keys()) < 1:
log.warning("No features are being tracked on " + server["server"]["address"])
# Clear previous totals
for tracked_feature in server["tracked_features"].values():
tracked_feature["usage_all"] = 0
tracked_feature["usage_nesi"] = 0
tracked_feature["users_nesi"] = {}
for featureorline in featureanduser_re_match:
group_dic = featureorline.groupdict()
if not group_dic["feature"] or not group_dic["user"]:
continue
# If this is the case, it is a feature header.
match_cluster = cluster_pattern.match(group_dic["host"])
# if duplicate, skip.
if group_dic["pid"]+group_dic["feature"] in pidpluslic_list:
log.debug("Duplicate licence entry.")
continue
else:
pidpluslic_list.append(group_dic["pid"]+group_dic["feature"])
log.debug(json.dumps(group_dic))
if group_dic["feature"] in server["tracked_features"].keys():
log.debug(group_dic["feature"] + " is tracked feature")
# Count gets added regardless of socket
if "count" in group_dic and group_dic["count"].isdigit():
lic_count = int(group_dic["count"])
else:
lic_count = 1
server["tracked_features"][group_dic["feature"]]["usage_all"] += lic_count
#print(server["tracked_features"][group_dic["feature"]]["usage_all"])
# Count gets added regardless of socket
if match_cluster:
#log.debug(group_dic["user"] + " is a user")
# If user not already. Add them.
if group_dic["user"] not in server["tracked_features"][group_dic["feature"]]["users_nesi"]:
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]] = {"count": 0, "tokens": 0, "sockets": [], "soak": 0}
server["tracked_features"][group_dic["feature"]]["usage_nesi"] += lic_count
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]]["count"] += lic_count
server["tracked_features"][group_dic["feature"]]["users_nesi"][group_dic["user"]]["sockets"].append(match_cluster.group(0))
elif match_cluster:
log.info("Untracked feature: " + group_dic["feature"] + " being used by " + group_dic["user"] + " on " + group_dic["host"])
elif group_dic["feature"] in server["untracked_features"]:
log.debug(group_dic["feature"] + " is untracked feature")
else:
server["untracked_features"].append(group_dic["feature"])
log.info("'" + group_dic["feature"] + "' being added to untracked features.")
log.debug(str(pidpluslic_list))
#shell_command_string = "export ANSYSLMD_LICENSE_FILE=$(head -n 1 " + server["licence_file"]["path"] + " | sed -n -e 's/.*=//p');linx64/ansysli_util -liusage"
# Read regex by line.
# Coarse
# for featureorline in featureanduser_re_match:
# group_dic = featureorline.groupdict()
# log.debug(json.dumps(group_dic))
# # If this is the case, it is a feature header.
# if group_dic["feature"] is not None:
# if group_dic["feature"] in server["tracked_features"].keys():
# log.debug(group_dic["feature"] + " is tracked feature")
# # Last_lic points to the licecne objects.
# if server["tracked_features"][group_dic["feature"]]["total"] != int(group_dic["total"]):
# log.warning("total was " + str(group_dic["total"]) + " at last count, now " + str(group_dic["total"]))
# server["tracked_features"][group_dic["feature"]]["total"] = int(group_dic["total"])
# server["tracked_features"][group_dic["feature"]]["usage_all"] = int(group_dic["count"])
# elif group_dic["feature"] in server["untracked_features"]:
# log.debug(group_dic["feature"] + " is untracked feature")
# else:
# server["untracked_features"].append(group_dic["feature"])
# log.info("'" + group_dic["feature"] + "' being added to untracked features.")
# Skip if disabled or non existant.
tic = time.time()
if "server" not in server or "polling_server" not in server["server"]:
log.warning("Skipping " + server["server"]["address"] + " as invalid details.")
server["server"] = settings["default"]["server"]
server["server"]["status"] = "INVALID"
return
if not server["server"]["polling_server"]:
log.info("Skipping server " + server["server"]["address"] + " as disabled.")
server["server"]["status"] = "DISABLED"
return
try:
tic = time.time()
server["server"]["status"] = "UNKNOWN"
locals()[server["server"]["poll_method"]]()
server["server"]["poll_time"] = (time.time()-tic)
except Exception as details:
log.error("Failed to check '" + server["server"]["address"] + "': " + str(type(details)) + " \n" + traceback.format_exc())
server["server"]["status"] = "DOWN"
else:
writemake_json(settings["path_store"], all_server_list)
schedul.enter(server["server"]["poll_period"], 1, poll_remote, argument=(server,))
log.debug(json.dumps(server))
def get_nesi_use():
log.info("Checking NeSI tokens... (period " + str(settings["squeue_poll_period"]) + "s)")
# Build a string of all licences to check.
all_licence_strings = {}
for server in server_list:
if "tracked_features" not in server:
continue # Skip if no features
for feature in server["tracked_features"].values():
if not feature["token_name"]:
continue
if not feature["slurm_active"]:
continue
for cluster in feature["clusters"]:
if cluster not in settings["clusters"].keys():
log.error("No such cluster " + cluster)
continue
if cluster not in all_licence_strings:
all_licence_strings[cluster] = ""
all_licence_strings[cluster] += feature["token_name"] + ","
log.debug(json.dumps(all_licence_strings))
# Return if nothing to check.
if not all_licence_strings:
return
# "clusters":{
# "mahuika":{
# "slurm_active":true
# },
# "maui":{
# },
# "maui_ancil":{}
# }
# For each cluster
for cluster, status in settings["clusters"].items():
if "enabled" not in status or not status["enabled"]:
log.info("Skipping cluster " + cluster + " disabled or missing details.")
continue
# Search squeue for running or pending jobs
sub_input = "squeue -h -M " + cluster + ' --format="%u|%C|%t|%r|%S|%N|%W" -t R -L ' + all_licence_strings[cluster]
# licence_pattern=re.compile(r"\s*(?P<username>\S*)\s*(?P<socket>\S*)\s*.*\), start (?P<datestr>.*?:.{2}).?\s?(?P<count>\d)?.*")
log.debug(sub_input)
try:
scontrol_string = ex_slurm_command(sub_input, "operator")
except Exception as details:
log.error("Failed to check scontrol licence usage. " + str(details))
else:
# Set current usage to zero
for server in server_list:
if "tracked_features" not in server:
continue
for feature in server["tracked_features"].values():
if "token_usage" not in feature:
continue
feature["token_usage"] = 0
# Read by line
scontrol_string_list = scontrol_string.split("\n")
scontrol_string_list.pop(0) # First Line is bleh
# try:
for line in scontrol_string_list:
#log.debug(line)
if len(line) < 6:
continue
line_delimited = line.split("|")
username = line_delimited[0]
licences_per_user = line_delimited[6].split(",")
# User may have multiple licences. Proccess for each.
for licence_token in licences_per_user:
if not licence_token:
continue
licence_token_name = licence_token.split(":")[0]
licence_token_count = licence_token.split(":")[1] if len(licence_token.split(":")) > 1 else 1
found = False
for server in server_list:
# Added flag because cant break out of two loops
for feature_name, feature_values in server["tracked_features"].items():
if feature_values["token_name"] == licence_token_name:
feature_values["token_usage"] += int(licence_token_count)
if username not in feature_values["users_nesi"]:
feature_values["users_nesi"][username] = {"count": 0, "tokens": 0, "sockets": [], "soak": 0}
feature_values["users_nesi"][username]["tokens"] += int(licence_token_count)
feature_values["users_nesi"][username]["soak"] -= int(licence_token_count)
found = True
if found:
break
if found:
break
else:
log.error("Licence " + licence_token_name + " does not exist in licence controller.")
log.info("Empty licence " + licence_token_name + " added to meta.")
schedul.enter(max(settings["squeue_poll_period"], 5), 1, get_nesi_use)
def validate():
"""Checks for inconinosistancies"""
if os.environ.get("VALIDATE", "").lower() == "false":
log.info("Skipping validation")
return
# Check resources recorded in sacctmgr
lic_ar={}
for line in ex_slurm_command("sacctmgr show resource withclusters -n -p", "operator").strip().split("\n"):
ls = line.split("|")
lic_ar[ls[0] + "@" + ls[1]] = ls
for server in server_list:
try:
for key, value in settings["default_server"].items():
if key not in server:
log.info(str(server) + " missing property '" + key + "'. Setting to default.")
server[key] = value
for feature, feature_values in server["tracked_features"].items():
for key, value in settings["default_feature"].items():
if key not in feature_values:
log.info(feature + " missing property '" + key + "'. Setting to default.")
feature_values[key] = value
# Copy dict name to value for backward comp.
if not feature_values["feature_name"]:
feature_values["feature_name"]=feature
# Compare with existing Tokens
clusters = feature_values["clusters"].copy()
num_clust=max(len(feature_values["clusters"]), 1)
fraction = int(100 / num_clust)
# 'total * clusters' IS NOT THE SAME AS 'total / (1/clusters)'
meta_total=math.ceil(feature_values["total"]/(fraction/100))
for token, values in lic_ar.items():
# List of clusters, remove once checked.
if token == feature_values["token_name"]:
if int(values[3]) != meta_total:
log.warning(token + " on " + values[6] + " has metatotal of " + str(values[3]) + " should have " + str(meta_total))
srmcmd="sacctmgr modify resource -i Name=" + values[0] + " Server=" + values[1] + " Set Count=" + str(meta_total)
log.error(srmcmd)
ex_slurm_command(srmcmd)
if int(values[4]) != fraction:
log.warning(token + " on " + values[6] + " has fraction of " + str(values[4]) + " should have " + str(fraction))
srmcmd="sacctmgr modify resource -i Name=" + values[0] + " Server=" + values[1] + " Clusters=" + values[6] + " Set PercentAllowed=" + str(fraction)
log.error(srmcmd)
ex_slurm_command(srmcmd)
# If token from cluster not in list.
if values[6] not in clusters:
log.warning("slurm licence token assigned on cluster " + values[6] + " but not in licence controller")
break
log.debug("Feature " + feature + " has cluster tokens " + ",".join(clusters) + " unnaccounted for.")
clusters.remove(values[6])
if clusters:
for cluster in clusters:
srmcmd="sacctmgr add resource -i Type=License Name=" + values[0] + " Server=" + values[1] + " Clusters=" + cluster + " Count=" + str(meta_total) + " PercentAllowed=" + str(fraction)
log.error(srmcmd)
#ex_slurm_command(srmcmd)
# if not feature_values["token_name"]:
# log.info(feature + " missing property '" + key + "'. Setting to default.")
# srvname=server["institution"]
# if server["faculty"]
# feature_values["token_name"]= feature.lower() + "@" + "faculty" "institution": "cwal219",
# Notify if no cluster set
if feature_values["slurm_active"] and feature_values["token_name"] and len(feature_values["clusters"]) < 1:
log.warning(feature_values["token_name"] + " is slurm_active, but is not assigned any cluster")
statdat = os.stat(server["licence_file"]["path"])
file_name = server["licence_file"]["path"].split("/")[-1]
owner = getpwuid(statdat.st_uid).pw_name
group = getgrgid(statdat.st_gid).gr_name
# Check permissions of file
if statdat.st_mode == 432:
raise Exception(server["licence_file"]["path"] + " file address permissions look weird.")
if server["licence_file"]["group"] and group != server["licence_file"]["group"]:
log.warning(server["licence_file"]["path"] + ' group is "' + group + '", should be "' + server["licence_file"]["group"] + '".')
if owner != settings["user"]:
log.warning(server["licence_file"]["path"] + " owner is '" + owner + "', should be '" + settings["user"] + "'.")
# if ll_value["licence_file_path"] != standard_address and ll_value["software_name"] and ll_value["institution"]:
# log.debug('Would be cool if "' + ll_value["licence_file_path"] + '" was "' + standard_address + '".')
# Read lic file contents
with open(server["licence_file"]["path"]) as file:
sub_out = file.read()
match_address = poll_methods[server["server"]["poll_method"]]["details_pattern"].match(sub_out).groupdict()
if not server["server"]["address"]:
server["server"]["address"] = match_address["server_address"]
elif server["server"]["address"] != match_address["server_address"]:
log.warning(file_name + " address mismatch: " + server["server"]["address"] + " -> " + match_address["server_address"])
if not server["server"]["port"]:
server["server"]["port"] = match_address["server_port"]
elif server["server"]["port"] != match_address["server_port"]:
log.warning(file_name + " port mismatch: " + server["server"]["port"] + " -> " + match_address["server_port"])
except Exception as details:
log.error("'" + server["licence_file"]["path"] + " is invalid: " + str(details))
server["server"]["polling_server"] = False
server["server"]["status"] = "INVALID"
writemake_json(settings["path_store"], all_server_list)
try:
settings = readmake_json("settings.json")
module_list = readmake_json(settings["path_modulelist"])
# Is correct user
if os.environ["USER"] != settings["user"] and not os.environ.get("CHECKUSER", "").lower() == "false":
log.error("Command should be run as '" + settings["user"] + "' as it owns licence files. ('export CHECKUSER=FALSE' to disable this check)")
exit()
# Clear
open("run_as_admin.sh", "w").close()
slurm_permissions = get_slurm_permssions()
# An error will be thrown if reservation is updated without change.
# Settings need to be fixed
log.debug(json.dumps(settings))
all_server_list = readmake_json(settings["path_store"])
server_list = []
if sys.argv[0] == os.path.basename(__file__):
sys.argv.pop(0)
if len(sys.argv) > 0:
name_list = ""
for arg in sys.argv:
found = False
for server in all_server_list:
if arg.strip() == server["software_name"]:
server_list.append(server)
with open(path) as json_file:
log.debug(path + " loaded.")
return json.load(json_file)
loop_start = time.time()
config = readmake_json("config.json", {"triggers": {}})
lazycache_username = readmake_json("cache/lazycache_username.json",
{"username_fullname": {}})
dir_parse = re.compile(
r"(?P<username>\S+)\s+(?P<directory>\S+)\s+(?P<available>((\d|\.)*[A-Z]|0)(?=\s*((\d|\.)*[A-Z]|0)))?\s+(?P<used>(\d|\.)*\D?)?\s+((?P<used_percent>(\d|\.)*)%)?\s+(?P<inodes>\S*)\s+(?P<inodes_used>\S+)\s+(?P<inodes_used_percent>\S+)%\n"
)
sub_input = "squeue --clusters all --state pending,running -h --format=%u | sort -u | sed -re '/(CLUSTER|USER|\s).*/d' | while read line; do sed -r -e '/Filesystem.*/d' -e 's/^/'\"$line\\t\"'/g' /scale_wlg_persistent/filesets/opt_nesi/var/quota/users/$line; done"
log.debug(sub_input)
output = subprocess.check_output(sub_input, shell=True).decode("utf-8")
log.debug(output)
dir_iterator = dir_parse.finditer(output.strip())
# Load cache
for trigger_name, trigger_val in config["triggers"].items():
trigger_val["cache"] = readmake_json("cache/" + trigger_name + ".json", {})
for directory in dir_iterator:
dir_dict = directory.groupdict("0")
if dir_dict["username"] in lazycache_username:
dir_dict["fullname"] = lazycache_username[dir_dict["username"]]
else:
log.debug(dir_dict["username"] +
" not in cache. Running ipa show-user")
