def register( self ):
self.set_node_timeout( 500 )
msgid = remap_utils.unique_id()
logger.info( "Registering with node" )
self.forward( "node._hello.%d"%(self.pid), {"msgid":msgid,"pid":self.pid,"priority":self.priority} )
# The while loop will terminate as soon as node stops sending messages,
# so this should be safe to do.
while True:
try:
msg = self.sub.recv()
msgprefix, data = remap_utils.unpack_msg( msg )
recipientid,msgtype,senderid = remap_utils.split_prefix(msgprefix)
if msgtype != "_hey":
continue
replymsgid = remap_utils.safe_get(data, "msgid")
if replymsgid == msgid:
# this is us
self.coreid = remap_utils.safe_get(data, "coreid" )
self.sub.set_string_option( nn.SUB, nn.SUB_UNSUBSCRIBE, "" )
self.sub.set_string_option( nn.SUB, nn.SUB_SUBSCRIBE, "global" )
self.sub.set_string_option( nn.SUB, nn.SUB_SUBSCRIBE, "local" )
self.sub.set_string_option( nn.SUB, nn.SUB_SUBSCRIBE, "notlocal" )
self.sub.set_string_option( nn.SUB, nn.SUB_SUBSCRIBE, self.coreid )
logger.info( "Received coreid %s."%( self.coreid ))
return True
except nn.NanoMsgAPIError as e:
logger.error( "Node is currently not available." )
break
logger.error( "Registration failed" )
return False
def process_hello( self, data ):
msgid = remap_utils.safe_get(data, "msgid")
pid = remap_utils.safe_get(data, "pid")
priority = remap_utils.safe_get( data, "priority" )
coreid = remap_utils.core_id( self.nodeid, pid )
self.cores[ coreid ] = {"coreid":coreid,"ts_last_seen":time.time(),"progress":-1,"pid":pid,"priority":priority}
msg = remap_utils.pack_msg( "%s._hey.%s"%(coreid, self.nodeid), {"msgid":msgid,"coreid":coreid} )
logger.info( "A core registered %s"%( coreid ))
self.lpub.send( msg )
def handle_jobstart( self, recipientid, senderid, data ):
avail, interruptable = self.hw.available_cpus( remap_utils.safe_get( data, "priority" ), self.cores )
numcores = len(remap_utils.safe_get( data, "cores" ))
if (avail + interruptable) >= numcores:
logger.info("Starting job with %d cores"%( numcores ))
if not self.hw.start_job( self.remaproot, senderid, numcores, data ):
logger.error("Error starting job")
else:
# Something changed in the meantime. Reject
logger.info( "Initiator requested %d cores, %d can be committed. Rejecting"%( numcores, avail + interruptable ))
msg = remap_utils.pack_msg( "%s.rejectjob.%s"%( senderid, self.nodeid ), {} )
self.forward_to_broker( msg )
self.coresChanged = True
def handle_jobstart(self, recipientid, senderid, data):
avail, interruptable = self.hw.available_cpus(
remap_utils.safe_get(data, "priority"), self.cores)
numcores = len(remap_utils.safe_get(data, "cores"))
if (avail + interruptable) >= numcores:
logger.info("Starting job with %d cores" % (numcores))
if not self.hw.start_job(self.remaproot, senderid, numcores, data):
logger.error("Error starting job")
else:
# Something changed in the meantime. Reject
logger.info(
"Initiator requested %d cores, %d can be committed. Rejecting"
% (numcores, avail + interruptable))
msg = remap_utils.pack_msg(
"%s.rejectjob.%s" % (senderid, self.nodeid), {})
self.forward_to_broker(msg)
self.coresChanged = True
def process_hello(self, data):
msgid = remap_utils.safe_get(data, "msgid")
pid = remap_utils.safe_get(data, "pid")
priority = remap_utils.safe_get(data, "priority")
coreid = remap_utils.core_id(self.nodeid, pid)
self.cores[coreid] = {
"coreid": coreid,
"ts_last_seen": time.time(),
"progress": -1,
"pid": pid,
"priority": priority
}
msg = remap_utils.pack_msg("%s._hey.%s" % (coreid, self.nodeid), {
"msgid": msgid,
"coreid": coreid
})
logger.info("A core registered %s" % (coreid))
self.lpub.send(msg)
def handle_showhands(self, recipientid, senderid, data):
avail, interruptable = self.hw.available_cpus(
remap_utils.safe_get(data, "priority"), self.cores)
if avail > 0 or interruptable > 0:
logger.info("Volunteering with %d cores, %d interruptable" %
(avail, interruptable))
msg = remap_utils.pack_msg("tracker.raisehand.%s" % (self.nodeid),
{
"free": avail,
"interruptable": interruptable
})
self.forward_to_broker(msg)
def handle_showhands( self, recipientid, senderid, data ):
avail, interruptable = self.hw.available_cpus( remap_utils.safe_get( data, "priority" ), self.cores )
if avail > 0 or interruptable > 0:
logger.info( "Volunteering with %d cores, %d interruptable"%( avail, interruptable ))
msg = remap_utils.pack_msg( "tracker.raisehand.%s"%( self.nodeid ), {"free":avail,"interruptable":interruptable} )
self.forward_to_broker( msg )