def show_resources(resources,
msg='Resources',
max_resources=None,
queues='default'):
"""Print the resources in a fancy way"""
if not max_resources:
max_resources = {}
total_hosts = 0
log = style.log_header(msg) + '\n'
for site in get_g5k_sites():
site_added = False
if site in resources:
log += style.log_header(site).ljust(20) + ' ' + str(
resources[site])
if site in max_resources:
log += '/' + str(max_resources[site])
log += ' '
site_added = True
for cluster in get_site_clusters(site, queues=queues):
if len(list(set(get_site_clusters(site)) & set(resources.keys()))) > 0 \
and not site_added:
log += style.log_header(site).ljust(20)
if site in max_resources:
log += '/' + str(max_resources[site])
log += ' '
site_added = True
if cluster in resources:
log += style.emph(cluster) + ': ' + str(resources[cluster])
if cluster in max_resources:
log += '/' + str(max_resources[cluster])
log += ' '
total_hosts += resources[cluster]
if site_added:
log += '\n'
if 'grid5000' in resources:
log += style.log_header('Grid5000').ljust(20) + str(
resources['grid5000'])
if "grid5000" in max_resources:
log += '/' + str(max_resources["grid5000"])
elif total_hosts > 0:
log += style.log_header('Total ').ljust(20) + str(total_hosts)
logger.info(log)
def __site_charter_remaining(site, day, user = None):
try:
with G5kAutoPortForwarder(site,
'oardb.' + site + '.grid5000.fr',
g5k_configuration['oar_pgsql_ro_port']) as (host, port):
start, end = get_oar_day_start_end(day)
if not user:
user = g5k_configuration.get('api_username')
if not user:
user = os.environ['LOGNAME']
conn = psycopg2.connect(host = host, port = port,
user = g5k_configuration['oar_pgsql_ro_user'],
password = g5k_configuration['oar_pgsql_ro_password'],
database = g5k_configuration['oar_pgsql_ro_db'])
try:
logger.trace("getting jobs for user %s on %s for %s" % (user, site, day))
OOC_total_site_used = 0
OOC_site_quota = 0
for cluster in get_site_clusters(site):
total_cluster_used = 0
for j in _get_jobs(conn, cluster, user, start, end):
logger.trace("%s:%s - job: start %s, end %s, walltime %s, %s" % (
site, cluster, format_unixts(j[7]),
format_unixts(j[8]), format_seconds(j[6]), j))
if _job_intersect_charter_period(j):
cluster_used = (j[9] + j[10]) * j[6]
logger.trace("%s:%s job %i intersects charter -> uses %is of cluster quota" % (
site, cluster, j[0], cluster_used,))
total_cluster_used += cluster_used
#cluster_quota = cluster_num_cores(cluster) * 3600 * 2
cluster_quota = _cluster_num_available_cores(conn, cluster) * 3600 * 2
logger.trace("%s:%s total cluster used = %i (%s), cluster quota = %i (%s)" % (
site, cluster,
total_cluster_used, format_seconds(total_cluster_used),
cluster_quota, format_seconds(cluster_quota)))
threading.currentThread().remaining[cluster] = max(0, cluster_quota - total_cluster_used)
OOC_total_site_used += total_cluster_used
OOC_site_quota += cluster_quota
logger.trace("%s to compare with outofchart: total used = %i (%s), %i%% of site quota = %i (%s)" % (
site,
OOC_total_site_used, format_seconds(OOC_total_site_used),
int(float(OOC_total_site_used) / float(OOC_site_quota) * 100.0),
OOC_site_quota, format_seconds(OOC_site_quota)))
finally:
conn.close()
except Exception as e:
logger.warn("error connecting to oar database / getting planning from " + site)
logger.detail("exception:\n" + format_exc())
def __site_charter_remaining(site, day, user = None):
try:
with G5kAutoPortForwarder(site,
'oardb.' + site + '.grid5000.fr',
g5k_configuration['oar_pgsql_ro_port']) as (host, port):
start, end = get_oar_day_start_end(day)
if not user:
user = g5k_configuration.get('api_username')
if not user:
user = os.environ['LOGNAME']
conn = psycopg2.connect(host = host, port = port,
user = g5k_configuration['oar_pgsql_ro_user'],
password = g5k_configuration['oar_pgsql_ro_password'],
database = g5k_configuration['oar_pgsql_ro_db'])
try:
logger.trace("getting jobs for user %s on %s for %s" % (user, site, day))
OOC_total_site_used = 0
OOC_site_quota = 0
for cluster in get_site_clusters(site):
total_cluster_used = 0
for j in _get_jobs(conn, cluster, user, start, end):
logger.trace("%s:%s - job: start %s, end %s, walltime %s, %s" % (
site, cluster, format_unixts(j[7]),
format_unixts(j[8]), format_seconds(j[6]), j))
if _job_intersect_charter_period(j):
cluster_used = (j[9] + j[10]) * j[6]
logger.trace("%s:%s job %i intersects charter -> uses %is of cluster quota" % (
site, cluster, j[0], cluster_used,))
total_cluster_used += cluster_used
#cluster_quota = cluster_num_cores(cluster) * 3600 * 2
cluster_quota = _cluster_num_available_cores(conn, cluster) * 3600 * 2
logger.trace("%s:%s total cluster used = %i (%s), cluster quota = %i (%s)" % (
site, cluster,
total_cluster_used, format_seconds(total_cluster_used),
cluster_quota, format_seconds(cluster_quota)))
threading.currentThread().remaining[cluster] = max(0, cluster_quota - total_cluster_used)
OOC_total_site_used += total_cluster_used
OOC_site_quota += cluster_quota
logger.trace("%s to compare with outofchart: total used = %i (%s), %i%% of site quota = %i (%s)" % (
site,
OOC_total_site_used, format_seconds(OOC_total_site_used),
int(float(OOC_total_site_used) / float(OOC_site_quota) * 100.0),
OOC_site_quota, format_seconds(OOC_site_quota)))
finally:
conn.close()
except Exception as e:
logger.warn("error connecting to oar database / getting planning from " + site)
logger.detail("exception:\n" + format_exc())
def _set_colors():
colors = {}
colors['busy'] = '#666666'
rgb_colors = [(x[0]/255., x[1]/255., x[2]/255.) for x in \
[(255., 122., 122.), (255., 204., 122.), (255., 255., 122.), (255., 246., 153.), (204., 255., 122.),
(122., 255., 122.), (122., 255., 255.), (122., 204., 255.), (204., 188., 255.), (255., 188., 255.)]]
i_site = 0
for site in sorted(get_g5k_sites()):
colors[site] = rgb_colors[i_site]
i_cluster = 0
for cluster in sorted(get_site_clusters(site, queues=None)):
min_index = colors[site].index(min(colors[site]))
color = [0., 0., 0.]
for i in range(3):
color[i] = min(colors[site][i], 1.)
if i == min_index:
color[i] += i_cluster * 0.12
colors[cluster] = tuple(color)
i_cluster += 1
i_site += 1
return colors
def get_vms_slot(vms, elements, slots, excluded_elements=None):
"""Return a slot with enough RAM and CPU """
chosen_slot = None
mem = vms[0]['mem']
cpu = vms[0]['n_cpu']
req_ram = sum([vm['mem'] for vm in vms])
req_cpu = sum([vm['n_cpu'] for vm in vms]) / 3
logger.debug('RAM %s CPU %s', req_ram, req_cpu)
for element in excluded_elements:
if element in get_g5k_sites():
excluded_elements += [
cluster for cluster in get_site_clusters(element)
if cluster not in excluded_elements
]
if 'grid5000' in elements:
clusters = [
cluster for cluster in get_g5k_clusters()
if cluster not in excluded_elements
and get_cluster_site not in excluded_elements
]
else:
clusters = [
element for element in elements if element in get_g5k_clusters()
and element not in excluded_elements
]
for element in elements:
if element in get_g5k_sites():
clusters += [
cluster for cluster in get_site_clusters(element)
if cluster not in clusters
and cluster not in excluded_elements
]
for slot in slots:
hosts = []
for element in slot[2]:
if str(element) in clusters:
n_hosts = slot[2][element]
for i in range(n_hosts):
hosts.append(Host(str(element + '-1.' + \
get_cluster_site(element) + '.grid5000.fr')))
attr = get_CPU_RAM_FLOPS(hosts)['TOTAL']
if attr['CPU'] > req_cpu and attr['RAM'] > req_ram:
chosen_slot = slot
break
del hosts[:]
if chosen_slot is None:
return None, None
resources_needed = {}
resources_available = chosen_slot[2]
logger.debug('resources available' + pformat(resources_available))
iter_clusters = cycle(clusters)
while req_ram > 0 or req_cpu > 0:
cluster = iter_clusters.next()
if resources_available[cluster] == 0:
clusters.remove(cluster)
iter_clusters = cycle(clusters)
else:
host = cluster + '-1'
attr = get_CPU_RAM_FLOPS([host])
resources_available[cluster] -= 1
req_ram -= float(attr[host]['RAM'] / mem) * mem
req_cpu -= float(attr[host]['CPU'] / cpu) * cpu
if cluster not in resources_needed:
resources_needed[cluster] = 0
resources_needed[cluster] += 1
if 'kavlan' in elements:
resources_needed['kavlan'] = 1
logger.debug('resources needed' + pformat(resources_needed))
return chosen_slot[0], distribute_hosts(chosen_slot[2], resources_needed,
excluded_elements)
def get_vms_slot(vms, elements, slots, excluded_elements=None):
"""Return a slot with enough RAM and CPU """
chosen_slot = None
mem = vms[0]['mem']
cpu = vms[0]['n_cpu']
req_ram = sum([vm['mem'] for vm in vms])
req_cpu = sum([vm['n_cpu'] for vm in vms]) / 3
logger.debug('RAM %s CPU %s', req_ram, req_cpu)
for element in excluded_elements:
if element in get_g5k_sites():
excluded_elements += [cluster for cluster
in get_site_clusters(element)
if cluster not in excluded_elements]
if 'grid5000' in elements:
clusters = [cluster for cluster in get_g5k_clusters()
if cluster not in excluded_elements
and get_cluster_site not in excluded_elements]
else:
clusters = [element for element in elements
if element in get_g5k_clusters()
and element not in excluded_elements]
for element in elements:
if element in get_g5k_sites():
clusters += [cluster
for cluster in get_site_clusters(element)
if cluster not in clusters
and cluster not in excluded_elements]
for slot in slots:
hosts = []
for element in slot[2]:
if str(element) in clusters:
n_hosts = slot[2][element]
for i in range(n_hosts):
hosts.append(Host(str(element + '-1.' + \
get_cluster_site(element) + '.grid5000.fr')))
attr = get_CPU_RAM_FLOPS(hosts)['TOTAL']
if attr['CPU'] > req_cpu and attr['RAM'] > req_ram:
chosen_slot = slot
break
del hosts[:]
if chosen_slot is None:
return None, None
resources_needed = {}
resources_available = chosen_slot[2]
logger.debug('resources available' + pformat(resources_available))
iter_clusters = cycle(clusters)
while req_ram > 0 or req_cpu > 0:
cluster = iter_clusters.next()
if resources_available[cluster] == 0:
clusters.remove(cluster)
iter_clusters = cycle(clusters)
else:
host = cluster + '-1'
attr = get_CPU_RAM_FLOPS([host])
resources_available[cluster] -= 1
req_ram -= float(attr[host]['RAM'] / mem) * mem
req_cpu -= float(attr[host]['CPU'] / cpu) * cpu
if cluster not in resources_needed:
resources_needed[cluster] = 0
resources_needed[cluster] += 1
if 'kavlan' in elements:
resources_needed['kavlan'] = 1
logger.debug('resources needed' + pformat(resources_needed))
return chosen_slot[0], distribute_hosts(chosen_slot[2], resources_needed,
excluded_elements)
def get_planning(elements=['grid5000'],
vlan=False,
subnet=False,
storage=False,
out_of_chart=False,
starttime=None,
endtime=None,
ignore_besteffort=True,
queues='default'):
"""Retrieve the planning of the elements (site, cluster) and others resources.
Element planning structure is ``{'busy': [(123456,123457), ... ], 'free': [(123457,123460), ... ]}.``
:param elements: a list of Grid'5000 elements ('grid5000', <site>, <cluster>)
:param vlan: a boolean to ask for KaVLAN computation
:param subnet: a boolean to ask for subnets computation
:param storage: a boolean to ask for sorage computation
:param out_of_chart: if True, consider that days outside weekends are busy
:param starttime: start of time period for which to compute the planning, defaults to now + 1 minute
:param endtime: end of time period for which to compute the planning, defaults to 4 weeks from now
:param ignore_besteffort: True by default, to consider the resources with besteffort jobs as available
:param queues: list of oar queues for which to get the planning
Return a dict whose keys are sites, whose values are dict whose keys
are cluster, subnets, kavlan or storage,
whose values are planning dicts, whose keys are hosts, subnet address range,
vlan number or chunk id planning respectively.
"""
if not starttime:
starttime = int(time() + timedelta_to_seconds(timedelta(minutes=1)))
starttime = int(get_unixts(starttime))
if not endtime:
endtime = int(starttime +
timedelta_to_seconds(timedelta(weeks=4, minutes=1)))
endtime = int(get_unixts(endtime))
if 'grid5000' in elements:
sites = elements = get_g5k_sites()
else:
sites = list(
set([site for site in elements if site in get_g5k_sites()] + [
get_cluster_site(cluster) for cluster in elements
if cluster in get_g5k_clusters(queues=queues)
] + [
get_host_site(host)
for host in elements if host in get_g5k_hosts()
or get_host_shortname(host) in get_g5k_hosts()
]))
if len(sites) == 0:
logger.error('Wrong elements given: %s' % (elements, ))
return None
planning = {}
for site in sites:
planning[site] = {}
for cluster in get_site_clusters(site, queues=queues):
planning[site][cluster] = {}
for site in sites:
if vlan:
planning[site].update({'vlans': {}})
if subnet:
planning[site].update({'subnets': {}})
if storage:
planning[site].update({'storage': {}})
if _retrieve_method == 'API':
_get_planning_API(planning, ignore_besteffort)
elif _retrieve_method == 'PostgreSQL':
_get_planning_PGSQL(planning, ignore_besteffort)
if out_of_chart:
_add_charter_to_planning(planning, starttime, endtime)
for site_pl in planning.values():
for res_pl in site_pl.values():
for el_planning in res_pl.values():
el_planning['busy'].sort()
_merge_el_planning(el_planning['busy'])
_trunc_el_planning(el_planning['busy'], starttime, endtime)
_fill_el_planning_free(el_planning, starttime, endtime)
# cleaning
real_planning = deepcopy(planning)
for site, site_pl in planning.items():
for cl, cl_pl in site_pl.items():
if cl in ['vlans']:
continue
keep_cluster = False
for h in cl_pl:
if not (get_host_site(h) in elements
or get_host_cluster(h) in elements
or get_host_shortname(h) in elements or h in elements):
del real_planning[site][cl][h]
else:
keep_cluster = True
if not keep_cluster:
del real_planning[site][cl]
return real_planning
def get_jobs_specs(resources, excluded_elements=None, name=None):
""" Generate the several job specifications from the dict of resources and
the blacklisted elements
:param resources: a dict, whose keys are Grid'5000 element and values the
corresponding number of n_nodes
:param excluded_elements: a list of elements that won't be used
:param name: the name of the jobs that will be given
"""
jobs_specs = []
if excluded_elements == None:
excluded_elements = []
# Creating the list of sites used
sites = []
real_resources = resources.copy()
for resource in resources:
if resource in get_g5k_sites() and resource not in sites:
sites.append(resource)
if resource in get_g5k_clusters(queues=None):
if resource not in excluded_elements:
site = get_cluster_site(resource)
if site not in sites:
sites.append(site)
if site not in real_resources:
real_resources[site] = 0
# Checking if we need a Kavlan, a KaVLAN global or none
get_kavlan = 'kavlan' in resources
if get_kavlan:
kavlan = 'kavlan'
n_sites = 0
for resource in real_resources:
if resource in sites:
n_sites += 1
if n_sites > 1:
kavlan += '-global'
break
blacklisted_hosts = {}
for element in excluded_elements:
if element not in get_g5k_clusters(queues=None) + get_g5k_sites():
site = get_host_site(element)
if not 'site' in blacklisted_hosts:
blacklisted_hosts[site] = [element]
else:
blacklisted_hosts[site].append(element)
for site in sites:
sub_resources = ''
# Adding a KaVLAN if needed
if get_kavlan:
if not 'global' in kavlan:
sub_resources = "{type='" + kavlan + "'}/vlan=1+"
get_kavlan = False
elif site in resources['kavlan']:
sub_resources = "{type='" + kavlan + "'}/vlan=1+"
get_kavlan = False
base_sql = '{'
end_sql = '}/'
# Creating blacklist SQL string for hosts
host_blacklist = False
str_hosts = ''
if site in blacklisted_hosts and len(blacklisted_hosts[site]) > 0:
str_hosts = ''.join([
"host not in ('" + get_host_longname(host) + "') and "
for host in blacklisted_hosts[site]
])
host_blacklist = True
#Adding the clusters blacklist
str_clusters = str_hosts if host_blacklist else ''
cl_blacklist = False
clusters_nodes = 0
for cluster in get_site_clusters(site, queues=None):
if cluster in resources and resources[cluster] > 0:
if str_hosts == '':
sub_resources += "{cluster='" + cluster + "'}"
else:
sub_resources += base_sql + str_hosts + "cluster='" + \
cluster + "'" + end_sql
sub_resources += "/nodes=" + str(resources[cluster]) + '+'
clusters_nodes += resources[cluster]
if cluster in excluded_elements:
str_clusters += "cluster not in ('" + cluster + "') and "
cl_blacklist = True
# Generating the site blacklist string from host and cluster blacklist
str_site = ''
if host_blacklist or cl_blacklist:
str_site += base_sql
if not cl_blacklist:
str_site += str_hosts[:-4]
else:
str_site += str_clusters[:-4]
str_site = str_site + end_sql
if real_resources[site] > 0:
sub_resources += str_site + "nodes=" + str(real_resources[site]) +\
'+'
if sub_resources != '':
jobs_specs.append((OarSubmission(resources=sub_resources[:-1],
name=name), site))
return jobs_specs