def dumpDetails(self):
pbs_server = pbs.pbs_default()
if not pbs_server:
print "No default pbs server"
sys.exit(1)
nodes = pbs.pbs_statnode(self.con, "", "NULL", "NULL")
for node in nodes:
print node.name
for attrib in node.attribs:
print '\t', attrib.name, '=', attrib.value
def getNodes(self):
'''
Build an array of Node objects comprising the worker nodes of the cluster.
'''
#pbs_statnode queries the pbs server over an existing connection and
#returns a list of nodes and some of their properties
nodelist = pbs.pbs_statnode(self.con, "", "NULL", "NULL")
#Iterate through the nodelist, creating
for node in nodelist:
thisnode = Node()
thisnode.setHostname(node.name)
thisnode.nodeType = 'hardware' #Node type is "hardware" unless it proves itself a cloud node!
##Bit of a hack, but set num jobs to zero initially - jobs data will not be given
# where there are no jobs running on a node, so overwrite this if that happens
thisnode.setNumJobs(0) ##Set number of running jobs to zero
thisnode.setFreeCpus(thisnode.num_cpus) #Set free cpus to the total number of cpus on the node
for attrib in node.attribs:
if attrib.name == 'state':
thisnode.setState(attrib.value)
elif attrib.name == 'np':
#np attribute contains the number of cpu cores
#as defined in Torques nodes file
thisnode.setNumCpus(attrib.value)
elif attrib.name == 'properties':
#"properties" is a resource-manager 'label' indicating
#any specific features provided by this node, such as
#applications or physical components
propertyList = attrib.value.split(',')
for propertyName in propertyList:
thisnode.addProperty(propertyName)
if propertyName == 'cloud': #Node has the "cloud" property
thisnode.nodeType = 'cloud' #Convert nodetype to cloud
#print "added property",propertyName
elif attrib.name == 'status':
#Torque 'status' contains a value which is in turn a string of attributes
#and corresponding values(e.g name1=value1,name2=value2 etc)
variables = attrib.value.split(',')
pairs = [variable.split('=',1) for variable in variables]
for data in pairs:
if data[0] == 'physmem':
thisnode.setMem(data[1])
elif data[0] == 'jobs':
if data[1]: ##if our list is not empty, this node jobs running on it.
for jobid in data[1]:
jobinstance = Job(jobid)
thisnode.addJob(jobinstance)
thisnode.setNumJobs(thisnode.jobs.__len__()) ##Add the number of running jobs as a property of the node
if thisnode.num_jobs == thisnode.num_cpus: ##If num_jobs = num_cpus, this node is completely full
thisnode.setFreeCpus(0)
self.fullnodes.append(thisnode) ##Add to list of full nodes
#else: print "data is ",data[0]
#else:
#print "attrib is",attrib.name,"and value is",attrib.value
#thisnode.printDetails()
self.nodes.append(thisnode)
if thisnode.num_jobs == 0 and thisnode.state != 'down':
thisnode.setState('idle')
if thisnode.state == 'down':
self.downnodes.append(thisnode)
elif thisnode.state == 'idle':
self.idlenodes.append(thisnode)
if thisnode.nodeType == 'cloud':
self.cloudnodes.append(thisnode)
if thisnode.state == 'idle':
self.idlecloudnodes.append(thisnode)
elif thisnode.nodeType == 'hardware':
if thisnode.state == 'idle':
self.idlehardwarenodes.append(thisnode)
self.total_nodes = len(self.nodes)
self.idle_nodes = len(self.idlenodes)
self.down_nodes = len(self.downnodes)
self.numIdleCloudNodes = len(self.idlecloudnodes)
self.numIdleHardwareNodes = len(self.idlehardwarenodes)