# along with vsc-jobs. If not, see <http://www.gnu.org/licenses/>.
# #
"""
This script reports jobs that are using more memory per core then the node average
@author: Stijn De Weirdt (Ghent University)
"""
from vsc.utils import fancylogger
from vsc.jobs.pbs.jobs import get_jobs_dict
from vsc.jobs.pbs.nodes import get_nodes_dict
_log = fancylogger.getLogger('show_mem')
ns = {}
for node, details in get_nodes_dict().items():
derived = details['derived']
if not 'np' in derived:
_log.warning('np not in derived from node %s' % node)
continue
np = derived['np']
if not 'physmem' in derived:
_log.warning('physmem not in derived from node %s' % node)
continue
mem = derived['physmem']
ns[node] = {
'np': np,
'mem': mem,
def main():
"""Main method"""
ns = {}
for node, details in get_nodes_dict().items():
derived = details['derived']
if not 'np' in derived:
_log.warning('np not in derived from node %s' % node)
continue
np = derived['np']
if not 'physmem' in derived:
_log.warning('physmem not in derived from node %s' % node)
continue
mem = derived['physmem']
ns[node] = {
'np': np,
'mem': mem,
'avg': int(mem / np),
}
toomuch = {}
MB = 1024 ** 2
for name, details in get_jobs_dict().items():
derived = details['derived']
if not derived['state'] in ('R',):
continue
if not 'user' in derived:
_log.warning("no user in derived job name %s" % name)
continue
user = derived['user']
if not 'used_mem' in derived:
_log.warning("no used_mem in derived job name %s" % name)
continue
used_mem = derived['used_mem']
if not 'exec_hosts' in derived:
_log.warning("no exec_hosts in derived job name %s" % name)
continue
exec_hosts = derived['exec_hosts']
cores = sum(exec_hosts.values())
used_avg_mem = int(used_mem / cores)
for host in exec_hosts.keys():
# more then avg on node?
if used_avg_mem > ns[host]['avg']:
if not user in toomuch:
toomuch[user] = {}
if not host in toomuch[user]:
toomuch[user][host] = []
toomuch[user][host].append([name, used_avg_mem / MB, ns[host]['avg'] / MB])
users = toomuch.keys()
users.sort()
txt = []
for user in users:
txt.append("%s:" % user)
for host, jobs in toomuch[user].items():
txt.append("\t%s:\t%s\n" % (host, ' '.join([str(x) for x in jobs])))
if len(txt) > 0:
print "\n".join(txt)
else:
print "All ok"
def main():
"""Main method"""
ns = {}
for node, details in get_nodes_dict().items():
derived = details['derived']
if not 'np' in derived:
_log.warning('np not in derived from node %s' % node)
continue
np = derived['np']
if not 'physmem' in derived:
_log.warning('physmem not in derived from node %s' % node)
continue
mem = derived['physmem']
ns[node] = {
'np': np,
'mem': mem,
'avg': int(mem / np),
}
toomuch = {}
MB = 1024**2
for name, details in get_jobs_dict().items():
derived = details['derived']
if not derived['state'] in ('R', ):
continue
if not 'user' in derived:
_log.warning("no user in derived job name %s" % name)
continue
user = derived['user']
if not 'used_mem' in derived:
_log.warning("no used_mem in derived job name %s" % name)
continue
used_mem = derived['used_mem']
if not 'exec_hosts' in derived:
_log.warning("no exec_hosts in derived job name %s" % name)
continue
exec_hosts = derived['exec_hosts']
cores = sum(exec_hosts.values())
used_avg_mem = int(used_mem / cores)
for host in exec_hosts.keys():
# more then avg on node?
if used_avg_mem > ns[host]['avg']:
if not user in toomuch:
toomuch[user] = {}
if not host in toomuch[user]:
toomuch[user][host] = []
toomuch[user][host].append(
[name, used_avg_mem / MB, ns[host]['avg'] / MB])
users = toomuch.keys()
users.sort()
txt = []
for user in users:
txt.append("%s:" % user)
for host, jobs in toomuch[user].items():
txt.append("\t%s:\t%s\n" % (host, ' '.join([str(x)
for x in jobs])))
if len(txt) > 0:
print "\n".join(txt)
else:
print "All ok"
