def print_VOViewLocal(queue_info, cp):
"""
Print out the VOView objects for the LSF batch system.
One VOView per VO per queue, for each VO which has access
to the queue.
"""
ce_name = cp.get(ce, "name")
vo_map = VoMapper(cp)
queue_jobs = getJobsInfo(vo_map, cp)
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
vo_queues = getVoQueues(queue_info, cp)
for vo, queue in vo_queues:
vo = vo.lower()
vo_info = queue_jobs.get(queue, {})
info2 = vo_info.get(vo, {})
ce_unique_id = buildCEUniqueID(cp, ce_name, 'lsf', queue)
my_queue_info = queue_info.setdefault(queue, {})
if cp.has_option("lsf", "max_wall"):
my_queue_info["max_wall"] = cp_getInt(cp, "lsf", "max_wall", 1440)
else:
if "max_wall" not in my_queue_info:
my_queue_info["max_wall"] = 1440
ert, wrt = responseTimes(cp,
info2.get("running", 0),
info2.get("waiting", 0),
max_job_time=my_queue_info.get("max_wall", 0))
free_slots = my_queue_info.get('free_slots', 0)
waiting = info2.get('waiting', 0)
if waiting > cp_getInt(cp, 'lsf', 'idle_slack', '10'):
free_slots = 0
info = {
'ceUniqueID': ce_unique_id,
'job_slots': my_queue_info.get('job_slots', 0),
'free_slots': free_slots,
'ce_name': ce_name,
'queue': queue,
'vo': vo,
'voLocalID': vo,
'job_manager': 'lsf',
'running': info2.get('running', 0),
'max_running': info2.get('max_running', 0),
'priority': queue_info.get(queue, {}).get('priority', 0),
'waiting': waiting,
'data': cp.get("osg_dirs", "data"),
'app': cp.get("osg_dirs", "app"),
'default_se': getDefaultSE(cp),
'ert': ert,
'wrt': wrt,
'acbr': 'VO:%s' % vo
}
info['total'] = info['waiting'] + info['running']
printTemplate(VOView, info)
def print_VOViewLocal(queue_info, cp):
"""
Print out the VOView objects for the LSF batch system.
One VOView per VO per queue, for each VO which has access
to the queue.
"""
ce_name = cp.get(ce, "name")
vo_map = VoMapper(cp)
queue_jobs = getJobsInfo(vo_map, cp)
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
vo_queues = getVoQueues(queue_info, cp)
for vo, queue in vo_queues:
vo = vo.lower()
vo_info = queue_jobs.get(queue, {})
info2 = vo_info.get(vo, {})
ce_unique_id = buildCEUniqueID(cp, ce_name, 'lsf', queue)
my_queue_info = queue_info.setdefault(queue, {})
if cp.has_option("lsf", "max_wall"):
my_queue_info["max_wall"] = cp_getInt(cp, "lsf", "max_wall", 1440)
else:
if "max_wall" not in my_queue_info:
my_queue_info["max_wall"] = 1440
ert, wrt = responseTimes(cp, info2.get("running", 0),
info2.get("waiting", 0),
max_job_time=my_queue_info.get("max_wall", 0))
free_slots = my_queue_info.get('free_slots', 0)
waiting = info2.get('waiting', 0)
if waiting > cp_getInt(cp, 'lsf', 'idle_slack', '10'):
free_slots = 0
info = {
'ceUniqueID' : ce_unique_id,
'job_slots' : my_queue_info.get('job_slots', 0),
'free_slots' : free_slots,
'ce_name' : ce_name,
'queue' : queue,
'vo' : vo,
'voLocalID' : vo,
'job_manager' : 'lsf',
'running' : info2.get('running', 0),
'max_running' : info2.get('max_running', 0),
'priority' : queue_info.get(queue, {}).get('priority', 0),
'waiting' : waiting,
'data' : cp.get("osg_dirs", "data"),
'app' : cp.get("osg_dirs", "app"),
'default_se' : getDefaultSE(cp),
'ert' : ert,
'wrt' : wrt,
'acbr' : 'VO:%s' % vo
}
info['total'] = info['waiting'] + info['running']
printTemplate(VOView, info)
def _generateSubClusterHelper(cp, section):
"""
Private helper function for generateSubClusters; do not use.
Implementation note: We create a bunch of variables but never
appear to use them - we actually do return the local dictionary
at the end of the function. Just being syntactically lazy.
"""
#pylint: disable-msg=W0612
# Names
subCluster = cp_get(cp, section, "name", cluster)
subClusterUniqueID = cp_get(cp, section, "unique_name", subCluster)
clusterUniqueID = getClusterID(cp)
# Host statistics
clockSpeed = cp_getInt(cp, section, "cpu_speed_mhz", 999999999)
cpuCount = cp_getInt(cp, section, "cpus_per_node", 2)
model = cp_get(cp, section, "cpu_model", 'UNDEFINEDVALUE')
platform = cp_get(cp, section, "platform", "UNKNOWN")
vendor = cp_get(cp, section, "cpu_vendor", 'UNDEFINEDVALUE')
cores_per_cpu = cp_getInt(cp, section, "cores_per_cpu", 2)
si2k = cp_getInt(cp, section, "SI00", 2000)
sf2k = cp_getInt(cp, section, "SF00", 2000)
hepspec = cp_getInt(cp, section, "HEPSPEC", 0)
ram = cp_getInt(cp, section, "ram_size", 1000*cpuCount*cores_per_cpu)
cores = cp_getInt(cp, section, "total_cores", 999999999)
if cores_per_cpu != 0:
cpus = cp_getInt(cp, section, "total_cpus", cores/cores_per_cpu)
else:
cpus = 0
virtualMem = ram + cp_getInt(cp, section, "swap_size", 0)
inboundIP = cp_getBoolean(cp, section, "inbound_network", False)
outboundIP = cp_getBoolean(cp, section, "outbound_network", True)
inboundIP = ldap_boolean(inboundIP)
outboundIP = ldap_boolean(outboundIP)
# OS Stats
osName, osRelease, osVersion = getRelease()
# Temp directories
default_tmp = cp_get(cp, "osg_dirs", "tmp", cp_get(cp, "osg_dirs", "data", \
"/tmp"))
wn_tmp = cp_get(cp, "osg_dirs", "wn_tmp", "/tmp")
tmp = cp_get(cp, section, "tmp", default_tmp)
if notDefined(tmp):
tmp = default_tmp
app_attr = 'GlueHostApplicationSoftwareRunTimeEnvironment'
apps = getApplications(cp)
applications = '\n'.join(['%s: %s' % (app_attr, i['locationId']) for i in \
apps if i['locationId']])
applications += '\n'
# BDII stuff
bdii = cp_get(cp, "bdii", "endpoint", "ldap://is.grid.iu.edu:2170")
return locals()
def main(cp = None, return_entries=False):
"""
Main method for the osg-info-wrapper script. This script safely runs the
plugin and provider modules, caching where necessary, and combines it with
the static data. It then outputs the final GLUE information for this site.
"""
log.debug("Starting up the osg-info-wrapper.")
if cp == None:
cp = config()
temp_dir = os.path.expandvars(cp_get(cp, "gip", "temp_dir", \
gipDir("$GIP_LOCATION/var/tmp", '/var/cache/gip')))
plugin_dir = os.path.expandvars(cp_get(cp, "gip", "plugin_dir", \
gipDir("$GIP_LOCATION/plugins", '/usr/libexec/gip/plugins')))
provider_dir = os.path.expandvars(cp_get(cp, "gip", "provider_dir", \
gipDir("$GIP_LOCATION/providers", '/usr/libexec/gip/providers')))
static_dir = os.path.expandvars(cp_get(cp, "gip", "static_dir", \
gipDir("$GIP_LOCATION/var/ldif", '/etc/gip/ldif.d')))
# Make sure that our directories exist.
create_if_not_exist(temp_dir, plugin_dir, provider_dir, static_dir)
# Load up our add, alter, and delete attributes
add_attributes = os.path.expandvars(cp_get(cp, "gip", \
"add_attributes", gipDir("$GIP_LOCATION/etc/add-attributes.conf",
'/etc/gip/add-attributes.conf')))
alter_attributes = os.path.expandvars(cp_get(cp, "gip", \
"alter_attributes", gipDir("$GIP_LOCATION/etc/alter-attributes.conf",
'/etc/gip/alter-attributes.conf')))
remove_attributes = os.path.expandvars(cp_get(cp, "gip", \
"remove_attributes", gipDir("$GIP_LOCATION/etc/remove-attributes.conf",
'/etc/gip/remove-attributes.conf')))
# Flush the cache if appropriate
do_flush_cache = cp_getBoolean(cp, "gip", "flush_cache", False)
if do_flush_cache:
log.info("Flushing cache upon request.")
flush_cache(temp_dir)
# Load up our parameters
freshness = cp_getInt(cp, "gip", "freshness", 300)
cache_ttl = cp_getInt(cp, "gip", "cache_ttl", 600)
response = cp_getInt(cp, "gip", "response", 240)
timeout = cp_getInt(cp, "gip", "timeout", 240)
try:
os.setpgrp()
except OSError, oe:
# If launched from a batch system (condor), we might not have perms
if oe.errno != 1:
raise
def main(cp=None, return_entries=False):
"""
Main method for the osg-info-wrapper script. This script safely runs the
plugin and provider modules, caching where necessary, and combines it with
the static data. It then outputs the final GLUE information for this site.
"""
log.debug("Starting up the osg-info-wrapper.")
if cp == None:
cp = config()
temp_dir = os.path.expandvars(cp_get(cp, "gip", "temp_dir", \
gipDir("$GIP_LOCATION/var/tmp", '/var/cache/gip')))
plugin_dir = os.path.expandvars(cp_get(cp, "gip", "plugin_dir", \
gipDir("$GIP_LOCATION/plugins", '/usr/libexec/gip/plugins')))
provider_dir = os.path.expandvars(cp_get(cp, "gip", "provider_dir", \
gipDir("$GIP_LOCATION/providers", '/usr/libexec/gip/providers')))
static_dir = os.path.expandvars(cp_get(cp, "gip", "static_dir", \
gipDir("$GIP_LOCATION/var/ldif", '/etc/gip/ldif.d')))
# Make sure that our directories exist.
create_if_not_exist(temp_dir, plugin_dir, provider_dir, static_dir)
# Load up our add, alter, and delete attributes
add_attributes = os.path.expandvars(cp_get(cp, "gip", \
"add_attributes", vdtDir("$VDT_LOCATION/gip/etc/add-attributes.conf",
'/etc/gip/add-attributes.conf')))
alter_attributes = os.path.expandvars(cp_get(cp, "gip", \
"alter_attributes", vdtDir("$VDT_LOCATION/gip/etc/alter-attributes.conf",
'/etc/gip/alter-attributes.conf')))
remove_attributes = os.path.expandvars(cp_get(cp, "gip", \
"remove_attributes", vdtDir("$VDT_LOCATION/gip/etc/remove-attributes.conf",
'/etc/gip/remove-attributes.conf')))
# Flush the cache if appropriate
do_flush_cache = cp_getBoolean(cp, "gip", "flush_cache", False)
if do_flush_cache:
log.info("Flushing cache upon request.")
flush_cache(temp_dir)
# Load up our parameters
freshness = cp_getInt(cp, "gip", "freshness", 300)
cache_ttl = cp_getInt(cp, "gip", "cache_ttl", 600)
response = cp_getInt(cp, "gip", "response", 240)
timeout = cp_getInt(cp, "gip", "timeout", 240)
try:
os.setpgrp()
except OSError, oe:
# If launched from a batch system (condor), we might not have perms
if oe.errno != 1:
raise
def print_VOViewLocal(queue_info, cp):
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
vo_map = VoMapper(cp)
queue_jobs = getJobsInfo(vo_map, cp)
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
vo_queues = getVoQueues(cp)
for vo, queue in vo_queues:
vo_info = queue_jobs.get(queue, {})
info2 = vo_info.get(vo, {})
port = getPort(cp)
ce_unique_id = buildCEUniqueID(cp, ce_name, 'pbs', queue)
my_queue_info = queue_info.setdefault(queue, {})
max_job_time = my_queue_info.get("max_wall", 0)
if cp.has_option("pbs", "max_wall"):
max_job_time = cp_getInt(cp, "pbs", "max_wall", 1440)
ert, wrt = responseTimes(cp, info2.get("running", 0),
info2.get("wait", 0),
max_job_time)
free_slots = my_queue_info.get('free_slots', 0)
waiting = info2.get('wait', 0)
if waiting > cp_getInt(cp, 'pbs', 'idle_slack', '10'):
free_slots = 0
info = {
'ceUniqueID' : ce_unique_id,
'job_slots' : my_queue_info.get('job_slots', 0),
'free_slots' : free_slots,
'ce_name' : ce_name,
'queue' : queue,
'vo' : vo,
'voLocalID' : vo,
'job_manager' : 'pbs',
'running' : info2.get('running', 0),
'max_running' : info2.get('max_running', 0),
'priority' : queue_info.get(queue, {}).get('priority', 0),
'waiting' : waiting,
'data' : cp_get(cp, "osg_dirs", "data", "UNKNOWN_DATA"),
'app' : cp_get(cp, "osg_dirs", "app", "UNKNOWN_APP"),
'default_se' : getDefaultSE(cp),
'ert' : 3600,
'wrt' : 3600,
'acbr' : 'VO:%s' % vo
}
info['total'] = info['waiting'] + info['running']
printTemplate(VOView, info)
def print_VOViewLocal(queue_info, cp):
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
vo_map = VoMapper(cp)
queue_jobs = getJobsInfo(vo_map, cp)
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
vo_queues = getVoQueues(cp)
for vo, queue in vo_queues:
vo_info = queue_jobs.get(queue, {})
info2 = vo_info.get(vo, {})
port = getPort(cp)
ce_unique_id = buildCEUniqueID(cp, ce_name, "pbs", queue)
my_queue_info = queue_info.setdefault(queue, {})
max_job_time = my_queue_info.get("max_wall", 0)
if cp.has_option("pbs", "max_wall"):
max_job_time = cp_getInt(cp, "pbs", "max_wall", 1440)
ert, wrt = responseTimes(cp, info2.get("running", 0), info2.get("wait", 0), max_job_time)
free_slots = my_queue_info.get("free_slots", 0)
waiting = info2.get("wait", 0)
if waiting > cp_getInt(cp, "pbs", "idle_slack", "10"):
free_slots = 0
info = {
"ceUniqueID": ce_unique_id,
"job_slots": my_queue_info.get("job_slots", 0),
"free_slots": free_slots,
"ce_name": ce_name,
"queue": queue,
"vo": vo,
"voLocalID": vo,
"job_manager": "pbs",
"running": info2.get("running", 0),
"max_running": info2.get("max_running", 0),
"priority": queue_info.get(queue, {}).get("priority", 0),
"waiting": waiting,
"data": cp_get(cp, "osg_dirs", "data", "UNKNOWN_DATA"),
"app": cp_get(cp, "osg_dirs", "app", "UNKNOWN_APP"),
"default_se": getDefaultSE(cp),
"ert": 3600,
"wrt": 3600,
"acbr": "VO:%s" % vo,
}
info["total"] = info["waiting"] + info["running"]
printTemplate(VOView, info)
def print_classic_access(cp, siteUniqueID):
"""
Emit the GLUE entity for a classic SE's access protocol.
"""
fallback_name = siteUniqueID + "_classicSE"
seUniqueID = cp_get(cp, "classic_se", "unique_name", fallback_name)
try:
default_host = socket.gethostname()
except:
default_host = 'UNKNOWN.example.org'
host = cp_get(cp, "classic_se", "host", default_host)
port = cp_getInt(cp, "classic_se", "port", "2811")
accessTemplate = getTemplate("GlueSE", "GlueSEAccessProtocolLocalID")
endpoint = 'gsiftp://%s:%i' % (host, port)
info = {'accessProtocolID' : 'GFTP_%s_%i' % (host, port),
'seUniqueID' : seUniqueID,
'protocol' : 'gsiftp',
'endpoint' : endpoint,
'capability' : 'file transfer',
'maxStreams' : 10,
'security' : 'gsiftp',
'port' : port,
'version' : '1.0.0',
}
print accessTemplate % info
def print_classic_access(cp, siteUniqueID):
"""
Emit the GLUE entity for a classic SE's access protocol.
"""
fallback_name = siteUniqueID + "_classicSE"
seUniqueID = cp_get(cp, "classic_se", "unique_name", fallback_name)
try:
default_host = socket.gethostname()
except:
default_host = 'UNKNOWN.example.org'
host = cp_get(cp, "classic_se", "host", default_host)
port = cp_getInt(cp, "classic_se", "port", "2811")
accessTemplate = getTemplate("GlueSE", "GlueSEAccessProtocolLocalID")
endpoint = 'gsiftp://%s:%i' % (host, port)
info = {
'accessProtocolID': 'GFTP_%s_%i' % (host, port),
'seUniqueID': seUniqueID,
'protocol': 'gsiftp',
'endpoint': endpoint,
'capability': 'file transfer',
'maxStreams': 10,
'security': 'gsiftp',
'port': port,
'version': '1.0.0',
}
print accessTemplate % info
def getSRMs(self):
"""
Return a list of dictionaries containing information about the SRM
endpoints.
Each dictionary must have the following keys:
- acbr
- status
- version
- endpoint
- name
The base class implementation uses the following configuration entries
(default value in parenthesis)
- se.srm_host (default: UNKNOWN.example.com)
- se.srm_version (2.2.0)
- se.srm_port (8443)
- se.srm_endpoint
(httpg://(se.srm_host):(se.srm_port)/srm/managerv2)
"""
srmname = cp_get(self._cp, self._section, "srm_host",
"UNKNOWN.example.com")
version = cp_get(self._cp, self._section, "srm_version", "2")
port = cp_getInt(self._cp, self._section, "srm_port", 8443)
if version.find('2') >= 0:
default_endpoint = 'httpg://%s:%i/srm/managerv2' % \
(srmname, int(port))
else:
default_endpoint = 'httpg://%s:%i/srm/managerv1' % \
(srmname, int(port))
endpoint = cp_get(self._cp, self._section, "srm_endpoint",
default_endpoint)
acbr_tmpl = '\nGlueServiceAccessControlRule: %s\n' \
'GlueServiceAccessControlRule: VO:%s'
acbr = ''
vos = voListStorage(self._cp, self._section)
for vo in vos:
acbr += acbr_tmpl % (vo, vo)
info = {'acbr': acbr[1:],
'status': 'OK',
'version': version,
'endpoint': endpoint,
'name': srmname,
}
return [info]
def getReferenceSI00(cp):
"""
The WLCG Installed Capacity document requires us to publish, for every CE,
the "reference" SI00 of a core for the CE. As we do not track the CE to
subcluster mapping, we just return the minimum SI00 of all the cluster
CPUs.
"""
global _referenceSI00
if _referenceSI00:
return _referenceSI00
_referenceSI00 = 2000
for section in cp.sections():
if not section.startswith(subcluster):
continue
_referenceSI00 = min(cp_getInt(cp, section, "SI00", 2000),
_referenceSI00)
return _referenceSI00
def getHTPCInfo(cp, batch, queue, log):
# return tuple: (non-Glue HTPC information, htpc_max_slots)
# where htpc_max_slots is the admin-provided "maximum number of slots per job"
htpcInfo = ('__GIP_DELETEME', 1) # defaults
if not cp_getBoolean(cp, batch, 'htpc_enabled', False):
log.info("HTPC is disabled for batch %s" % batch)
return htpcInfo
log.info("HTPC is enabled for batch %s" % batch)
whitelist = cp_getList(cp, batch, 'htpc_queues', [])
blacklist = cp_getList(cp, batch, 'htpc_blacklist_queues', [])
log.debug("HTPC whitelist: %s; HTPC blacklist %s: " % (whitelist, blacklist))
if '*' not in whitelist and queue not in whitelist:
log.info("HTPC Queue %s not in whitelist" % queue)
return htpcInfo
if queue in blacklist:
log.info("HTPC Queue %s in blacklist" % queue)
return htpcInfo
defaultRSL = cp_get(cp, batch, 'htpc_rsl', '')
log.debug("HTPC DefaultRSL: %s" % defaultRSL)
queueRSL = cp_get(cp, batch, 'htpc_rsl_%s' % queue, '')
if not queueRSL:
queueRSL = defaultRSL
if not queueRSL:
log.info("HTPC RSL not found for queue %s" % queue)
return htpcInfo
htpcMaxSlots = cp_getInt(cp, batch, 'htpc_max_slots', 1)
if htpcMaxSlots < 2:
log.info("HTPC max slots equal to 1 or not set!")
return htpcInfo
# acbr stuff?
return ('HTPCrsl: %s' % queueRSL, htpcMaxSlots)
def print_CE(cp):
"""
Print out the GlueCE objects for LSF; one GlueCE per grid queue.
"""
try:
lsfVersion = getLrmsInfo(cp)
except:
lsfVersion = 'Unknown'
log.debug('Using LSF version %s' % lsfVersion)
queueInfo = getQueueInfo(cp)
try:
totalCpu, freeCpu, queueCpus = parseNodes(queueInfo, cp)
except:
#raise
totalCpu, freeCpu, queueCpus = 0, 0, {}
log.debug('Total, Free CPU: (%s, %s)' % (totalCpu, freeCpu))
ce_name = cp.get(ce, "name")
CE = getTemplate("GlueCE", "GlueCEUniqueID")
try:
excludeQueues = [i.strip() for i in cp.get("lsf", \
"queue_exclude").split(',')]
except:
excludeQueues = []
vo_queues = getVoQueues(queueInfo, cp)
for queue, info in queueInfo.items():
if queue in excludeQueues:
continue
log.debug('Processing queue %s' % queue)
if 'running' not in info:
info['running'] = 0
if 'status' not in info:
# There really should be an unknown status...
info['status'] = 'Closed'
if 'total' not in info:
info['total'] = 0
info["lrmsVersion"] = lsfVersion
info["job_manager"] = "lsf"
if int(info.get("wait", 0)) > 0:
info["free_slots"] = 0
else:
if queue in queueCpus and 'max' in queueCpus[queue] and 'njobs' in queueCpus[queue]:
info["free_slots"] = queueCpus[queue]['max'] - queueCpus[queue]['njobs']
else:
info["free_slots"] = freeCpu
info["queue"] = queue
info["ceName"] = ce_name
unique_id = buildCEUniqueID(cp, ce_name, 'lsf', queue)
info['ceUniqueID'] = unique_id
if "job_slots" not in info:
if queue in queueCpus and 'max' in queueCpus[queue]:
log.debug('queue %s, info is %s' % (queue, queueCpus[queue]))
info['job_slots'] = queueCpus[queue]['max']
else:
info["job_slots"] = totalCpu
if "priority" not in info:
info["priority"] = 0
if "max_running" not in info:
info["max_running"] = info["job_slots"]
elif not info['max_running'] or info['max_running'] == '-':
info['max_running'] = 999999
if cp.has_option("lsf", "max_wall"):
info["max_wall"] = cp_getInt(cp, "lsf", "max_wall", 1440)
else:
if "max_wall" not in info:
info["max_wall"] = 1440
info["max_wall"] = int(info["max_wall"]) # glue proscribes ints
info["job_slots"] = min(totalCpu, info["job_slots"])
ert, wrt = responseTimes(cp, info["running"], info["wait"],
max_job_time=info["max_wall"])
contact_string = buildContactString(cp, 'lsf', queue, unique_id, log)
ceImpl, ceImplVersion = getCEImpl(cp)
info['ert'] = ert
info['wrt'] = wrt
info['hostingCluster'] = cp_get(cp, ce, 'hosting_cluster', ce_name)
info['hostName'] = cp_get(cp, ce, 'host_name', ce_name)
info['ceImpl'] = ceImpl
info['ceImplVersion'] = ceImplVersion
info['contact_string'] = contact_string
info['app_dir'] = cp.get('osg_dirs', 'app')
info['data_dir'] = cp.get('osg_dirs', 'data')
info['default_se'] = getDefaultSE(cp)
info['max_waiting'] = 999999
#info['max_total'] = info['max_running']
info['max_total'] = info['max_waiting'] + info['max_running']
info['assigned'] = info['job_slots']
info['lrmsType'] = 'lsf'
info['preemption'] = str(cp_getInt(cp, 'lsf', 'preemption', '0'))
acbr = ''
for vo, queue2 in vo_queues:
if queue == queue2:
acbr += 'GlueCEAccessControlBaseRule: VO:%s\n' % vo.lower()
if not acbr:
continue
#print info
info['acbr'] = acbr[:-1]
info['bdii'] = cp.get('bdii', 'endpoint')
gramVersion = getGramVersion(cp)
port = getPort(cp)
info['gramVersion'] = gramVersion
info['port'] = port
info['waiting'] = info.get('wait', 0)
info['referenceSI00'] = gip_cluster.getReferenceSI00(cp)
info['clusterUniqueID'] = getClusterID(cp)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
htpcRSL, maxSlots = getHTPCInfo(cp, 'lsf', queue, log)
info['max_slots'] = maxSlots
info['htpc'] = htpcRSL
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
info['extraCapabilities'] = extraCapabilities
printTemplate(CE, info)
return queueInfo, totalCpu, freeCpu, queueCpus
def print_VOViewLocal(cp):
"""
Print the GLUE VOView entity; shows the VO's view of the condor batch
system.
Config options used:
* ce.name. The human-readable name of the ce.
* condor.status. The status of condor; defaults to "Production"
* osg_dirs.app. The $OSG_APP directory; defaults to "/Unknown"
* osg_dirs.data. The $OSG_DATA directory; defaults to "/Unknown"
* se.name. The human-readable name of the closest SE.
@param cp: The GIP configuration object
@type cp: ConfigParser.ConfigParser
"""
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
ce_name = cp_get(cp, "ce", "name", "")
#status = cp_get(cp, "condor", "status", "Production")
#condorVersion = getLrmsInfo(cp)
total_nodes, _, unclaimed = parseNodes(cp)
vo_map = VoMapper(cp)
jobs_info = getJobsInfo(vo_map, cp)
groupInfo = getGroupInfo(vo_map, cp)
# Add in the default group
all_group_vos = []
total_assigned = 0
for key, val in groupInfo.items():
if key == 'default':
continue
all_group_vos.extend(val['vos'])
total_assigned += val.get('quota', 0)
all_vos = sets.Set(voList(cp))
defaultVoList = [i for i in all_vos if i not in all_group_vos]
if 'default' not in groupInfo:
groupInfo['default'] = {}
groupInfo['default']['vos'] = defaultVoList
if total_nodes > total_assigned:
log.info("There are %i assigned job slots out of %i total; assigning" \
" the rest to the default group." % (total_assigned, total_nodes))
groupInfo['default']['quota'] = total_nodes - total_assigned
else:
log.warning("More assigned nodes (%i) than actual nodes (%i)!" % \
(total_assigned, total_nodes))
if defaultGroupIsExcluded(cp):
if groupInfo.has_key('default'):
del groupInfo['default']
for group in groupInfo:
jinfo = jobs_info.get(group, {})
vos = sets.Set(groupInfo[group].get('vos', [group]))
vos.update(jinfo.keys())
vos.intersection_update(all_vos)
# Enforce invariants
# VO_FREE_SLOTS <= CE_FREE_SLOTS
# VO_FREE_SLOTS <= CE_ASSIGNED - VO_RUNNING
# This code determines CE_ASSIGNED
ginfo = groupInfo[group]
if ginfo.get("quota", 0) > 0:
assigned = ginfo.get("quota", 0)
else:
assigned = total_nodes
log.debug("All VOs for %s: %s" % (group, ", ".join(vos)))
ce_unique_id = buildCEUniqueID(cp, ce_name, 'condor', group)
max_wall = cp_getInt(cp, "condor", "max_wall", 1440)
myrunning = sum([i.get('running', 0) for i in jinfo.values()], 0)
assigned = max(assigned, myrunning)
for vo in vos:
acbr = 'VO:%s' % vo
info = jinfo.get(vo.lower(), {"running": 0, "idle": 0, "held": 0})
ert, wrt = responseTimes(cp, info["running"], info["idle"] + \
info["held"], max_job_time=max_wall*60)
free = min(unclaimed, assigned - myrunning,
assigned - int(info['running']))
free = int(free)
waiting = int(info["idle"]) + int(info["held"])
if waiting > cp_getInt(cp, 'condor', 'idle_slack', '10'):
free = 0
info = {
"vo": vo,
"acbr": acbr,
"ceUniqueID": ce_unique_id,
"voLocalID": vo,
"ce_name": ce_name,
"job_manager": 'condor',
"queue": vo,
"running": info["running"],
# Held jobs are included as "waiting" since the definition is:
# Number of jobs that are in a state different than running
"waiting": waiting,
"total": info["running"] + info["idle"] + info["held"],
"free_slots": free,
"job_slots": int(total_nodes),
"ert": ert,
"wrt": wrt,
"default_se": getDefaultSE(cp),
'app': cp_get(cp, 'osg_dirs', 'app', '/Unknown'),
"data": cp_get(cp, "osg_dirs", "data", "/Unknown"),
}
printTemplate(VOView, info)
# If the running jobs are greater than the total/assigned, bump
# up the values of the total/assigned
# Keeps the invariant: RUNNING <= ASSIGNED, RUNNING <= TOTAL
assigned = max(assigned, myrunning)
ce_total_nodes = max(assigned, ce_total_nodes)
# Make sure the following holds:
# CE_FREE_SLOTS <= ASSIGNED - RUNNING
# CE_FREE_SLOTS <= UNCLAIMED SLOTS IN CONDOR
ce_unclaimed = min(assigned - myrunning, unclaimed)
myidle = sum([i.get('idle', 0) for i in jinfo.values()], 0)
myheld = sum([i.get('held', 0) for i in jinfo.values()], 0)
max_wall = cp_getInt(cp, "condor", "max_wall", 1440)
ert, wrt = responseTimes(cp,
myrunning,
myidle + myheld,
max_job_time=max_wall * 60)
referenceSI00 = gip_cluster.getReferenceSI00(cp)
contact_string = buildContactString(cp, 'condor', group, ce_unique_id,
log)
htpcRSL, maxSlots = getHTPCInfo(cp, 'condor', group, log)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
def print_VOViewLocal(cp):
"""
Print the GLUE VOView entity; shows the VO's view of the condor batch
system.
Config options used:
* ce.name. The human-readable name of the ce.
* condor.status. The status of condor; defaults to "Production"
* osg_dirs.app. The $OSG_APP directory; defaults to "/Unknown"
* osg_dirs.data. The $OSG_DATA directory; defaults to "/Unknown"
* se.name. The human-readable name of the closest SE.
@param cp: The GIP configuration object
@type cp: ConfigParser.ConfigParser
"""
VOView = getTemplate("GlueCE", "GlueVOViewLocalID")
ce_name = cp_get(cp, "ce", "name", "")
#status = cp_get(cp, "condor", "status", "Production")
#condorVersion = getLrmsInfo(cp)
total_nodes, _, unclaimed = parseNodes(cp)
vo_map = VoMapper(cp)
jobs_info = getJobsInfo(vo_map, cp)
groupInfo = getGroupInfo(vo_map, cp)
# Add in the default group
all_group_vos = []
total_assigned = 0
for key, val in groupInfo.items():
if key == 'default':
continue
all_group_vos.extend(val['vos'])
total_assigned += val.get('quota', 0)
all_vos = sets.Set(voList(cp))
defaultVoList = [i for i in all_vos if i not in all_group_vos]
if 'default' not in groupInfo:
groupInfo['default'] = {}
groupInfo['default']['vos'] = defaultVoList
if total_nodes > total_assigned:
log.info("There are %i assigned job slots out of %i total; assigning" \
" the rest to the default group." % (total_assigned, total_nodes))
groupInfo['default']['quota'] = total_nodes-total_assigned
else:
log.warning("More assigned nodes (%i) than actual nodes (%i)!" % \
(total_assigned, total_nodes))
if defaultGroupIsExcluded(cp):
if groupInfo.has_key('default'):
del groupInfo['default']
for group in groupInfo:
jinfo = jobs_info.get(group, {})
vos = sets.Set(groupInfo[group].get('vos', [group]))
vos.update(jinfo.keys())
vos.intersection_update(all_vos)
# Enforce invariants
# VO_FREE_SLOTS <= CE_FREE_SLOTS
# VO_FREE_SLOTS <= CE_ASSIGNED - VO_RUNNING
# This code determines CE_ASSIGNED
ginfo = groupInfo[group]
if ginfo.get("quota", 0) > 0:
assigned = ginfo.get("quota", 0)
else:
assigned = total_nodes
log.debug("All VOs for %s: %s" % (group, ", ".join(vos)))
ce_unique_id = buildCEUniqueID(cp, ce_name, 'condor', group)
max_wall = cp_getInt(cp, "condor", "max_wall", 1440)
myrunning = sum([i.get('running', 0) for i in jinfo.values()], 0)
assigned = max(assigned, myrunning)
for vo in vos:
acbr = 'VO:%s' % vo
info = jinfo.get(vo.lower(), {"running": 0, "idle": 0, "held": 0})
ert, wrt = responseTimes(cp, info["running"], info["idle"] + \
info["held"], max_job_time=max_wall*60)
free = min(unclaimed, assigned-myrunning,
assigned-int(info['running']))
free = int(free)
waiting = int(info["idle"]) + int(info["held"])
if waiting > cp_getInt(cp, 'condor', 'idle_slack', '10'):
free = 0
info = {"vo" : vo,
"acbr" : acbr,
"ceUniqueID" : ce_unique_id,
"voLocalID" : vo,
"ce_name" : ce_name,
"job_manager" : 'condor',
"queue" : vo,
"running" : info["running"],
# Held jobs are included as "waiting" since the definition is:
# Number of jobs that are in a state different than running
"waiting" : waiting,
"total" : info["running"] + info["idle"] + info["held"],
"free_slots" : free,
"job_slots" : int(total_nodes),
"ert" : ert,
"wrt" : wrt,
"default_se" : getDefaultSE(cp),
'app' : cp_get(cp, 'osg_dirs', 'app', '/Unknown'),
"data" : cp_get(cp, "osg_dirs", "data", "/Unknown"),
}
printTemplate(VOView, info)
# If the running jobs are greater than the total/assigned, bump
# up the values of the total/assigned
# Keeps the invariant: RUNNING <= ASSIGNED, RUNNING <= TOTAL
assigned = max(assigned, myrunning)
ce_total_nodes = max(assigned, ce_total_nodes)
# Make sure the following holds:
# CE_FREE_SLOTS <= ASSIGNED - RUNNING
# CE_FREE_SLOTS <= UNCLAIMED SLOTS IN CONDOR
ce_unclaimed = min(assigned - myrunning, unclaimed)
myidle = sum([i.get('idle', 0) for i in jinfo.values()], 0)
myheld = sum([i.get('held', 0) for i in jinfo.values()], 0)
max_wall = cp_getInt(cp, "condor", "max_wall", 1440)
ert, wrt = responseTimes(cp, myrunning, myidle+myheld,
max_job_time=max_wall*60)
referenceSI00 = gip_cluster.getReferenceSI00(cp)
contact_string = buildContactString(cp, 'condor', group, ce_unique_id, log)
htpcRSL, maxSlots = getHTPCInfo(cp, 'condor', group, log)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
def print_CE(cp):
pbsVersion = getLrmsInfo(cp)
queueInfo = getQueueInfo(cp)
totalCpu, freeCpu, queueCpus = parseNodes(cp, pbsVersion)
log.debug("totalCpu, freeCpu, queueCPus: %s %s %s" % (totalCpu, freeCpu, queueCpus))
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
CE = getTemplate("GlueCE", "GlueCEUniqueID")
try:
excludeQueues = [i.strip() for i in cp_get(cp, "pbs", "queue_exclude", "").split(",")]
except:
excludeQueues = []
vo_queues = getVoQueues(cp)
for queue, info in queueInfo.items():
if queue in excludeQueues:
continue
info["lrmsVersion"] = pbsVersion
info["job_manager"] = "pbs"
# if no jobs are waiting in the queue, set the number of free slots
# to (job_slots - running), or the total number of free slots on the cluster,
# whichever is less.
info["queue"] = queue
info["ceName"] = ce_name
unique_id = buildCEUniqueID(cp, ce_name, "pbs", queue)
ceImpl, ceImplVersion = getCEImpl(cp)
port = getPort(cp)
info["ceUniqueID"] = unique_id
if "job_slots" not in info:
info["job_slots"] = totalCpu
if "priority" not in info:
info["priority"] = 0
if "max_running" not in info:
info["max_running"] = info["job_slots"]
if cp.has_option("pbs", "max_wall"):
info["max_wall"] = cp_getInt(cp, "pbs", "max_wall", 1440)
else:
if "max_wall" not in info:
info["max_wall"] = 1440
info["free_slots"] = 0
if info["wait"] == 0:
freeSlots = info["job_slots"] - info["running"]
if freeSlots > 0:
info["free_slots"] = min(freeSlots, freeCpu)
log.debug("queue info: %s %s" % (queue, info))
ert, wrt = responseTimes(cp, info.get("running", 0), info.get("wait", 0), max_job_time=info["max_wall"])
info["job_slots"] = min(totalCpu, info["job_slots"])
info["ert"] = ert
info["wrt"] = wrt
info["hostingCluster"] = cp_get(cp, ce, "hosting_cluster", ce_name)
info["hostName"] = cp_get(cp, ce, "host_name", ce_name)
info["ceImpl"] = ceImpl
info["ceImplVersion"] = ceImplVersion
contact_string = buildContactString(cp, "pbs", queue, unique_id, log)
info["contact_string"] = contact_string
info["app_dir"] = cp_get(cp, "osg_dirs", "app", "/UNKNOWN_APP")
info["data_dir"] = cp_get(cp, "osg_dirs", "data", "/UNKNOWN_DATA")
info["default_se"] = getDefaultSE(cp)
if "max_waiting" not in info:
info["max_waiting"] = 999999
if "max_queuable" in info:
info["max_total"] = info["max_queuable"]
info["free_slots"] = min(info["free_slots"], info["max_queuable"])
else:
info["max_total"] = info["max_waiting"] + info["max_running"]
info["free_slots"] = min(info["free_slots"], info["max_total"])
# Enforce invariants:
# max_total <= max_running
# free_slots <= max_running
info["max_total"] = min(info["max_total"], info["max_running"])
info["free_slots"] = min(info["free_slots"], info["max_running"])
info["assigned"] = info["job_slots"]
# Enforce invariants:
# assigned <= max_running
info["assigned"] = min(info["assigned"], info["max_running"])
info["lrmsType"] = "pbs"
info["preemption"] = cp_get(cp, "pbs", "preemption", "0")
acbr = ""
has_vo = False
for vo, queue2 in vo_queues:
if queue == queue2:
acbr += "GlueCEAccessControlBaseRule: VO:%s\n" % vo
has_vo = True
if not has_vo:
continue
info["acbr"] = acbr[:-1]
info["bdii"] = cp.get("bdii", "endpoint")
gramVersion = getGramVersion(cp)
info["gramVersion"] = gramVersion
info["port"] = port
info["waiting"] = info["wait"]
info["referenceSI00"] = gip_cluster.getReferenceSI00(cp)
info["clusterUniqueID"] = getClusterID(cp)
extraCapabilities = ""
if cp_getBoolean(cp, "site", "glexec_enabled", False):
extraCapabilities = extraCapabilities + "\n" + "GlueCECapability: glexec"
htpcRSL, maxSlots = getHTPCInfo(cp, "pbs", queue, log)
info["max_slots"] = maxSlots
if maxSlots > 1:
extraCapabilities = extraCapabilities + "\n" + "GlueCECapability: htpc"
info["extraCapabilities"] = extraCapabilities
info["htpc"] = htpcRSL
printTemplate(CE, info)
return queueInfo, totalCpu, freeCpu, queueCpus
def print_CE(cp):
SGEVersion = getLrmsInfo(cp)
queueInfo, _ = getQueueInfo(cp)
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
ce_template = getTemplate("GlueCE", "GlueCEUniqueID")
queueList = getQueueList(cp)
vo_queues = getVoQueues(cp)
default_max_waiting = 999999
for queue in queueInfo.values():
if 'name' not in queue or queue['name'] not in queueList:
continue
if queue['name'] == 'waiting':
continue
unique_id = buildCEUniqueID(cp, ce_name, 'sge', queue['name'])
acbr = ''
for vo, queue2 in vo_queues:
if queue['name'] == queue2:
acbr += 'GlueCEAccessControlBaseRule: VO:%s\n' % vo
referenceSI00 = gip_cluster.getReferenceSI00(cp)
contact_string = buildContactString(cp, 'sge', queue['name'],
unique_id, log)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
htpcRSL, maxSlots = getHTPCInfo(cp, 'sge', queue, log)
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
gramVersion = getGramVersion(cp)
port = getPort(cp)
ceImpl, ceImplVersion = getCEImpl(cp)
max_wall = queue["max_wall"]
if cp.has_option("sge", "max_wall"):
max_wall = cp_getInt(cp, "sge", "max_wall", 1440)
info = { \
"ceUniqueID" : unique_id,
"ceName" : ce_name,
"ceImpl" : ceImpl,
"ceImplVersion" : ceImplVersion,
"clusterUniqueID" : getClusterID(cp),
"queue" : queue['name'],
"priority" : queue['priority'],
"lrmsType" : 'sge',
"lrmsVersion" : SGEVersion,
"job_manager" : "sge",
"job_slots" : queue["slots_total"],
"free_slots" : queue["slots_free"],
"running" : queue["slots_used"],
"status" : queue['status'],
"total" : queue['slots_used'] + queue['waiting'],
"ert" : 3600,
"wrt" : 3600,
"hostingCluster" : cp_get(cp, ce, 'hosting_cluster', ce_name),
"hostName" : cp_get(cp, ce, 'host_name', ce_name),
"contact_string" : contact_string,
"app_dir" : cp_get(cp, 'osg_dirs', 'app', "/OSG_APP_UNKNOWN"),
"data_dir" : cp_get(cp, 'osg_dirs', 'data', "/OSG_DATA_UNKNOWN"),
"default_se" : getDefaultSE(cp),
"max_running" : queue["slots_total"],
"max_wall" : max_wall,
"max_waiting" : default_max_waiting,
"max_slots" : maxSlots,
"max_total" : default_max_waiting + queue["slots_total"],
"assigned" : queue["slots_used"],
"preemption" : cp_get(cp, 'sge', 'preemption', '0'),
"acbr" : acbr[:-1],
"bdii": cp.get('bdii', 'endpoint'),
"gramVersion" : gramVersion,
"port" : port,
"waiting" : queue['waiting'],
"referenceSI00": referenceSI00,
'extraCapabilities' : extraCapabilities,
"htpc" : htpcRSL
}
printTemplate(ce_template, info)
return queueInfo
def print_CE(cp):
"""
Print out the GlueCE objects for LSF; one GlueCE per grid queue.
"""
try:
lsfVersion = getLrmsInfo(cp)
except:
lsfVersion = 'Unknown'
log.debug('Using LSF version %s' % lsfVersion)
queueInfo = getQueueInfo(cp)
try:
totalCpu, freeCpu, queueCpus = parseNodes(queueInfo, cp)
except:
#raise
totalCpu, freeCpu, queueCpus = 0, 0, {}
log.debug('Total, Free CPU: (%s, %s)' % (totalCpu, freeCpu))
ce_name = cp.get(ce, "name")
CE = getTemplate("GlueCE", "GlueCEUniqueID")
try:
excludeQueues = [i.strip() for i in cp.get("lsf", \
"queue_exclude").split(',')]
except:
excludeQueues = []
vo_queues = getVoQueues(queueInfo, cp)
for queue, info in queueInfo.items():
if queue in excludeQueues:
continue
log.debug('Processing queue %s' % queue)
if 'running' not in info:
info['running'] = 0
if 'status' not in info:
# There really should be an unknown status...
info['status'] = 'Closed'
if 'total' not in info:
info['total'] = 0
info["lrmsVersion"] = lsfVersion
info["job_manager"] = "lsf"
if int(info.get("wait", 0)) > 0:
info["free_slots"] = 0
else:
if queue in queueCpus and 'max' in queueCpus[
queue] and 'njobs' in queueCpus[queue]:
info["free_slots"] = queueCpus[queue]['max'] - queueCpus[
queue]['njobs']
else:
info["free_slots"] = freeCpu
info["queue"] = queue
info["ceName"] = ce_name
unique_id = buildCEUniqueID(cp, ce_name, 'lsf', queue)
info['ceUniqueID'] = unique_id
if "job_slots" not in info:
if queue in queueCpus and 'max' in queueCpus[queue]:
log.debug('queue %s, info is %s' % (queue, queueCpus[queue]))
info['job_slots'] = queueCpus[queue]['max']
else:
info["job_slots"] = totalCpu
if "priority" not in info:
info["priority"] = 0
if "max_running" not in info:
info["max_running"] = info["job_slots"]
elif not info['max_running'] or info['max_running'] == '-':
info['max_running'] = 999999
if cp.has_option("lsf", "max_wall"):
info["max_wall"] = cp_getInt(cp, "lsf", "max_wall", 1440)
else:
if "max_wall" not in info:
info["max_wall"] = 1440
info["max_wall"] = int(info["max_wall"]) # glue proscribes ints
info["job_slots"] = min(totalCpu, info["job_slots"])
ert, wrt = responseTimes(cp,
info["running"],
info["wait"],
max_job_time=info["max_wall"])
contact_string = buildContactString(cp, 'lsf', queue, unique_id, log)
ceImpl, ceImplVersion = getCEImpl(cp)
info['ert'] = ert
info['wrt'] = wrt
info['hostingCluster'] = cp_get(cp, ce, 'hosting_cluster', ce_name)
info['hostName'] = cp_get(cp, ce, 'host_name', ce_name)
info['ceImpl'] = ceImpl
info['ceImplVersion'] = ceImplVersion
info['contact_string'] = contact_string
info['app_dir'] = cp.get('osg_dirs', 'app')
info['data_dir'] = cp.get('osg_dirs', 'data')
info['default_se'] = getDefaultSE(cp)
info['max_waiting'] = 999999
#info['max_total'] = info['max_running']
info['max_total'] = info['max_waiting'] + info['max_running']
info['assigned'] = info['job_slots']
info['lrmsType'] = 'lsf'
info['preemption'] = str(cp_getInt(cp, 'lsf', 'preemption', '0'))
acbr = ''
for vo, queue2 in vo_queues:
if queue == queue2:
acbr += 'GlueCEAccessControlBaseRule: VO:%s\n' % vo.lower()
if not acbr:
continue
#print info
info['acbr'] = acbr[:-1]
info['bdii'] = cp.get('bdii', 'endpoint')
gramVersion = getGramVersion(cp)
port = getPort(cp)
info['gramVersion'] = gramVersion
info['port'] = port
info['waiting'] = info.get('wait', 0)
info['referenceSI00'] = gip_cluster.getReferenceSI00(cp)
info['clusterUniqueID'] = getClusterID(cp)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
htpcRSL, maxSlots = getHTPCInfo(cp, 'lsf', queue, log)
info['max_slots'] = maxSlots
info['htpc'] = htpcRSL
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
info['extraCapabilities'] = extraCapabilities
printTemplate(CE, info)
return queueInfo, totalCpu, freeCpu, queueCpus
def print_CE(cp):
SGEVersion = getLrmsInfo(cp)
queueInfo, _ = getQueueInfo(cp)
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
ce_template = getTemplate("GlueCE", "GlueCEUniqueID")
queueList = getQueueList(cp)
vo_queues = getVoQueues(cp)
default_max_waiting = 999999
for queue in queueInfo.values():
if 'name' not in queue or queue['name'] not in queueList:
continue
if queue['name'] == 'waiting':
continue
unique_id = buildCEUniqueID(cp, ce_name, 'sge', queue['name'])
acbr = ''
for vo, queue2 in vo_queues:
if queue['name'] == queue2:
acbr += 'GlueCEAccessControlBaseRule: VO:%s\n' % vo
referenceSI00 = gip_cluster.getReferenceSI00(cp)
contact_string = buildContactString(cp, 'sge', queue['name'], unique_id, log)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
htpcRSL, maxSlots = getHTPCInfo(cp, 'sge', queue, log)
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
gramVersion = getGramVersion(cp)
port = getPort(cp)
ceImpl, ceImplVersion = getCEImpl(cp)
max_wall = queue["max_wall"]
if cp.has_option("sge", "max_wall"):
max_wall = cp_getInt(cp, "sge", "max_wall", 1440)
info = { \
"ceUniqueID" : unique_id,
"ceName" : ce_name,
"ceImpl" : ceImpl,
"ceImplVersion" : ceImplVersion,
"clusterUniqueID" : getClusterID(cp),
"queue" : queue['name'],
"priority" : queue['priority'],
"lrmsType" : 'sge',
"lrmsVersion" : SGEVersion,
"job_manager" : "sge",
"job_slots" : queue["slots_total"],
"free_slots" : queue["slots_free"],
"running" : queue["slots_used"],
"status" : queue['status'],
"total" : queue['slots_used'] + queue['waiting'],
"ert" : 3600,
"wrt" : 3600,
"hostingCluster" : cp_get(cp, ce, 'hosting_cluster', ce_name),
"hostName" : cp_get(cp, ce, 'host_name', ce_name),
"contact_string" : contact_string,
"app_dir" : cp_get(cp, 'osg_dirs', 'app', "/OSG_APP_UNKNOWN"),
"data_dir" : cp_get(cp, 'osg_dirs', 'data', "/OSG_DATA_UNKNOWN"),
"default_se" : getDefaultSE(cp),
"max_running" : queue["slots_total"],
"max_wall" : max_wall,
"max_waiting" : default_max_waiting,
"max_slots" : maxSlots,
"max_total" : default_max_waiting + queue["slots_total"],
"assigned" : queue["slots_used"],
"preemption" : cp_get(cp, 'sge', 'preemption', '0'),
"acbr" : acbr[:-1],
"bdii": cp.get('bdii', 'endpoint'),
"gramVersion" : gramVersion,
"port" : port,
"waiting" : queue['waiting'],
"referenceSI00": referenceSI00,
'extraCapabilities' : extraCapabilities,
"htpc" : htpcRSL
}
printTemplate(ce_template, info)
return queueInfo
def print_CE(cp):
pbsVersion = getLrmsInfo(cp)
queueInfo = getQueueInfo(cp)
totalCpu, freeCpu, queueCpus = parseNodes(cp, pbsVersion)
log.debug("totalCpu, freeCpu, queueCPus: %s %s %s" % (totalCpu, freeCpu, queueCpus))
ce_name = cp_get(cp, ce, "name", "UNKNOWN_CE")
CE = getTemplate("GlueCE", "GlueCEUniqueID")
try:
excludeQueues = [i.strip() for i in cp_get(cp, "pbs", \
"queue_exclude", "").split(',')]
except:
excludeQueues = []
vo_queues = getVoQueues(cp)
for queue, info in queueInfo.items():
if queue in excludeQueues:
continue
info["lrmsVersion"] = pbsVersion
info["job_manager"] = "pbs"
# if no jobs are waiting in the queue, set the number of free slots
# to (job_slots - running), or the total number of free slots on the cluster,
# whichever is less.
info["queue"] = queue
info["ceName"] = ce_name
unique_id = buildCEUniqueID(cp, ce_name, 'pbs', queue)
ceImpl, ceImplVersion = getCEImpl(cp)
port = getPort(cp)
info['ceUniqueID'] = unique_id
if "job_slots" not in info:
info["job_slots"] = totalCpu
if "priority" not in info:
info["priority"] = 0
if "max_running" not in info:
info["max_running"] = info["job_slots"]
if cp.has_option("pbs", "max_wall"):
info["max_wall"] = cp_getInt(cp, "pbs", "max_wall", 1440)
else:
if "max_wall" not in info:
info["max_wall"] = 1440
info["free_slots"] = 0
if info["wait"] == 0:
freeSlots = info["job_slots"] - info["running"]
if freeSlots > 0:
info["free_slots"] = min(freeSlots, freeCpu)
log.debug("queue info: %s %s" % (queue, info))
ert, wrt = responseTimes(cp, info.get("running", 0),
info.get("wait", 0), max_job_time=info["max_wall"])
info["job_slots"] = min(totalCpu, info["job_slots"])
info['ert'] = ert
info['wrt'] = wrt
info['hostingCluster'] = cp_get(cp, ce, 'hosting_cluster', ce_name)
info['hostName'] = cp_get(cp, ce, 'host_name', ce_name)
info['ceImpl'] = ceImpl
info['ceImplVersion'] = ceImplVersion
contact_string = buildContactString(cp, 'pbs', queue, unique_id, log)
info['contact_string'] = contact_string
info['app_dir'] = cp_get(cp, 'osg_dirs', 'app', "/UNKNOWN_APP")
info['data_dir'] = cp_get(cp, 'osg_dirs', 'data', "/UNKNOWN_DATA")
info['default_se'] = getDefaultSE(cp)
if 'max_waiting' not in info:
info['max_waiting'] = 999999
if 'max_queuable' in info:
info['max_total'] = info['max_queuable']
info['free_slots'] = min(info['free_slots'], info['max_queuable'])
else:
info['max_total'] = info['max_waiting'] + info['max_running']
info['free_slots'] = min(info['free_slots'], info['max_total'])
# Enforce invariants:
# max_total <= max_running
# free_slots <= max_running
info['max_total'] = min(info['max_total'], info['max_running'])
info['free_slots'] = min(info['free_slots'], info['max_running'])
info['assigned'] = info['job_slots']
# Enforce invariants:
# assigned <= max_running
info['assigned'] = min(info['assigned'], info['max_running'])
info['lrmsType'] = 'pbs'
info['preemption'] = cp_get(cp, 'pbs', 'preemption', '0')
acbr = ''
has_vo = False
for vo, queue2 in vo_queues:
if queue == queue2:
acbr += 'GlueCEAccessControlBaseRule: VO:%s\n' % vo
has_vo = True
if not has_vo:
continue
info['acbr'] = acbr[:-1]
info['bdii'] = cp.get('bdii', 'endpoint')
gramVersion = getGramVersion(cp)
info['gramVersion'] = gramVersion
info['port'] = port
info['waiting'] = info['wait']
info['referenceSI00'] = gip_cluster.getReferenceSI00(cp)
info['clusterUniqueID'] = getClusterID(cp)
extraCapabilities = ''
if cp_getBoolean(cp, 'site', 'glexec_enabled', False):
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: glexec'
htpcRSL, maxSlots = getHTPCInfo(cp, 'pbs', queue, log)
info['max_slots'] = maxSlots
if maxSlots > 1:
extraCapabilities = extraCapabilities + '\n' + 'GlueCECapability: htpc'
info['extraCapabilities'] = extraCapabilities
info['htpc'] = htpcRSL
printTemplate(CE, info)
return queueInfo, totalCpu, freeCpu, queueCpus