def setup_cluster():
print("setting up simple cluster")
setup = ClusterSetup('simple')
setup.conf = {
'cors': {
'authorized_origins': 'http://10.5.0.1:8080',
},
'config': {
'authentication': 'enable',
'racksxml': '/etc/slurm-web/racks.xml',
},
'roles': {
'guest': 'enabled',
'all': 'enabled',
'user': '******',
'admin': '@admin',
'restricted_fields_for_all': 'command',
'restricted_fields_for_user': '******',
'restricted_fields_for_admin': '',
},
'acl': {},
'ldap': {
'uri': 'ldap://localhost:ldap',
'base_people': 'ou=people,dc=cluster,dc=local',
'base_group': 'ou=group,dc=cluster,dc=local',
'expiration': '1296000',
},
'cache': {
'redis_host': '',
'redis_port': '',
'jobs_expiration': '',
'global_expiration': '',
}
}
setup.racks = MockRacksXML()
setup.racks.add_nodetype('nodetype1', 'model type 1', '1', '1')
row = setup.racks.add_racksrow('0')
rack = setup.racks.add_rack(row, 'rack1', '0')
setup.racks.add_nodeset(rack, 'cn[01-30]', 'nodetype1', '0', '2')
setup.userbase = UserBase()
setup.userbase.add(User('pierre', 'curie', 'toto', ['user', 'admin']))
setup.userbase.add(User('marie', 'curie', 'toto', ['user', 'admin']))
setup.ctld = SlurmCtld(u'cluster1')
for nodeid in range(1, 30):
nodename = u"cn%02d" % (nodeid)
node = SlurmNode(nodename)
if nodeid == 3:
node.state = 'MIXED'
setup.ctld.nodes.add(node)
partition = SlurmPartition('compute')
partition.nodes = 'cn[01-30]'
partition.total_nodes = 30
partition.total_cpus = 8 * 30
partition.total_mem = 1 * 30
setup.ctld.partitions.add(partition)
xqos = SlurmQos('qos_test')
xqos.grp_jobs = 10
xqos.max_tres_pj = u'1=224,4=8'
xqos.priority = 100
setup.ctld.qos.add(xqos)
job = SlurmJob('1234')
job.name = 'test.sh'
job.qos = xqos.name
job.job_state = 'RUNNING'
job.nodes = 'cn03'
job.num_nodes = 1
job.num_cpus = 1
job.start_time = int(time.time() - 10)
job.time_limit = 3600
job.cpus_allocated = {'cn3': 1}
job.group_id = setup.userbase[0].gid
job.user_id = setup.userbase[0].uid
job.account = 'physic'
job.shared = 2 ^ 16 - 2
job.work_dir = u'/home/pierre'
job.command = u'/home/pierre/sleep.sh'
job.partition = partition.name
setup.ctld.jobs.add(job)
return setup
def setup_cluster():
name = u"jupiter"
print("setting up cluster %s" % (name))
setup = ClusterSetup(name)
setup.conf = {
'cors': {
'authorized_origins': 'http://10.5.0.1:8080',
},
'config': {
'authentication': 'enable',
'racksxml': '/etc/slurm-web/racks.xml',
},
'roles': {
'guests': 'enabled',
'all': 'disabled',
'user': '******',
'admin': '@admin',
'restricted_fields_for_all': 'command',
'restricted_fields_for_user': '******',
'restricted_fields_for_admin': '',
},
'acl': {},
'ldap': {
'uri': 'ldaps://localhost:ldaps',
'base_people': 'ou=people,dc=cluster,dc=local',
'base_group': 'ou=group,dc=cluster,dc=local',
'expiration': '1296000',
'cacert': '/path/to/cert'
},
'cache': {
'redis_host': '',
'redis_port': '',
'jobs_expiration': '',
'global_expiration': '',
}
}
setup.racks = MockRacksXML()
setup.racks.add_nodetype('blade', 'blade x2', '4', str(1. / 5))
#setup.racks.add_nodetype('blade', 'blade x2', '1', str(1./2))
row = setup.racks.add_racksrow('0')
rack = setup.racks.add_rack(row, 'A', '0')
setup.racks.add_nodeset(rack, 'cn[001-040]', 'blade', '0', '2')
rack = setup.racks.add_rack(row, 'B', '0')
setup.racks.add_nodeset(rack, 'cn[041-080]', 'blade', '0', '2')
rack = setup.racks.add_rack(row, 'C', '0')
setup.racks.add_nodeset(rack, 'cn[081-120]', 'blade', '0', '2')
rack = setup.racks.add_rack(row, 'D', '0')
setup.racks.add_nodeset(rack, 'cn[121-160]', 'blade', '0', '2')
rack = setup.racks.add_rack(row, 'E', '0')
setup.racks.add_nodeset(rack, 'cn[161-200]', 'blade', '0', '2')
rack = setup.racks.add_rack(row, 'F', '0')
setup.racks.add_nodeset(rack, 'cn[201-240]', 'blade', '0', '2')
setup.userbase = UserBase()
setup.userbase.add(User('pierre', 'curie', 'toto', ['user', 'admin']))
setup.userbase.add(User('marie', 'curie', 'toto', ['user', 'admin']))
setup.userbase.add(User('jacques', 'curie', 'toto', ['user']))
setup.userbase.add(User('thomas', 'curie', 'toto', []))
setup.ctld = SlurmCtld(name)
nodeset = NodeSet('cn[001-240]')
for nodename in nodeset:
node = SlurmNode(nodename)
node.cpus = 24
node.cores = node.cpus / node.sockets
setup.ctld.nodes.add(node)
partition = SlurmPartition('compute')
partition.nodes = str(nodeset)
partition.total_nodes = len(nodeset)
partition.total_cpus = 10 * partition.total_nodes
partition.total_mem = 1 * partition.total_nodes
setup.ctld.partitions.add(partition)
switch = SlurmTopology('swibmaster1')
switch.switches = 'swibleaf[1-8]'
switch.nodes = 'cn[001-240]'
switch.level = 1
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibmaster2')
switch.switches = 'swibleaf[1-8]'
switch.nodes = 'cn[001-240]'
switch.level = 1
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf1')
switch.nodes = 'cn[001-030]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf2')
switch.nodes = 'cn[031-060]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf3')
switch.nodes = 'cn[061-090]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf4')
switch.nodes = 'cn[091-120]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf5')
switch.nodes = 'cn[121-150]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf6')
switch.nodes = 'cn[151-180]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf7')
switch.nodes = 'cn[181-210]'
setup.ctld.topology.add(switch)
switch = SlurmTopology('swibleaf8')
switch.nodes = 'cn[211-240]'
setup.ctld.topology.add(switch)
xqos = SlurmQos('qos_test')
xqos.grp_jobs = 10
xqos.max_tres_pj = u'1=224,4=8'
xqos.priority = 100
setup.ctld.qos.add(xqos)
first_id = 1235
# get the number of cpus from the first node
nb_cores = list(setup.ctld.nodes)[0].cpus
for jobid in range(first_id, first_id + 1000):
nodename = nodeset[int((jobid - first_id) / nb_cores)]
node = setup.ctld.find_node(nodename)
#cpu = node.alloc_cpus # pick up the first not allocated cpu
job = SlurmJob(str(jobid))
job.name = 'test.sh'
job.qos = xqos.name
job.job_state = 'RUNNING'
job.nodes = nodename
job.num_nodes = 1
job.num_cpus = 1
job.start_time = int(time.time() - 10)
job.time_limit = 3600
job.cpus_allocated = {nodename: 1}
job.group_id = setup.userbase[0].gid
job.user_id = setup.userbase[0].uid
job.account = 'physic'
job.shared = 2 ^ 16 - 2
job.work_dir = u'/home/pierre'
job.command = u'/home/pierre/test.sh'
job.partition = partition.name
setup.ctld.jobs.add(job)
node.alloc_cpus += 1
if node.alloc_cpus == node.cpus:
node.state = u"ALLOCATED"
else:
node.state = u"MIXED"
return setup