def txn():
index = random.randint(0, NUM_SHARDS - 1)
shard_name = "shard" + str(index)
counter = SSFSMSimpleCounterShard.get_by_key_name(shard_name)
if counter is None:
counter = SSFSMSimpleCounterShard(key_name=shard_name)
counter.count += value
counter.put()
def fsm_calculate_run_time():
""" Fantasm does not give call backs when its done. Must figure it out
with another job using the last modified date on output entities
"""
# Get the last job which was run for subset /fsm
q = Record.all()
q.filter('engine_type =','fsm')
q.filter('benchmark =','subset')
q.order('-start')
results = q.fetch(1)
if len(results) == 0:
logging.error("Unable to find a record for fsm/subset")
return False
q = None
record = None
for ii in results:
if ii.state == "Done":
logging.error("Last FSM end time has already been calculated")
logging.info(str(ii.num_entities))
q = SSFSMSimpleCounterShard.all()
if not q:
logging.error("No query returned for SubSet results")
return False
record = ii
max_date = None
while True:
results = q.fetch(1000)
for ii in results:
date = ii.modified
if max_date == None or max_date < date:
max_date = date
if len(results) < 1000:
break;
if not max_date:
logging.error("Unable to calculate the max date for FSM/subset")
return False
record.state = "Done"
record.end = max_date
delta = (record.end - record.start)
record.total = float(delta.days * 86400 + delta.seconds) + float(delta.microseconds)/1000000
record.put()
return True
def post(self):
if self.request.get("fsm_cleanup"):
if fsm_calculate_run_time():
self.redirect("/subset")
else:
self.response.out.write("Error calculating run time of FSM/subset")
if self.request.get("reset_fsm_count"):
for c in SSFSMSimpleCounterShard.all():
c.delete()
self.redirect('/subset')
return
if self.request.get("reset_mr_count"):
for c in SSMRSimpleCounterShard.all():
c.delete()
self.redirect('/subset')
return
if self.request.get("reset_pl_count"):
for c in SSPLSimpleCounterShard.all():
c.delete()
self.redirect('/subset')
return
if self.request.get("compute"):
engine = self.request.get("engine")
dataset = self.request.get("dataset")
user = self.request.get('user')
data = SubSetDataSet.get_by_key_name(dataset)
record = Record(engine_type=engine,
dataset=dataset,
benchmark="subset",
num_entities=data.num_entries,
entries_per_pipe=data.entries_per_pipe,
user=user,
state="Running")
if engine == "fsm":
record.put()
# reset count
for c in SSFSMSimpleCounterShard.all():
c.delete()
context = {}
context['user'] = str(user)
context['num_entries'] = int(data.num_entries)
fsm.startStateMachine('SubSet', [context])
self.redirect('/subset')
elif engine == "pipeline":
for c in SSPLSimpleCounterShard.all():
c.delete()
mypipeline = SubSetPipeline(data.num_entries)
mypipeline.start()
record.pipeline_id = mypipeline.pipeline_id
record.put()
self.redirect('/subset')
#self.redirect(mypipeline.base_path + "/status?root=" + mypipeline.pipeline_id)
elif engine == "mr":
for c in SSMRSimpleCounterShard.all():
c.delete()
# Why 1k each per shard or less? is this ideal?
if data.num_entries > 1000: shards = data.num_entries/1000
else: shards = 1
kind = get_class(data.num_entries)
mapreduce_id = control.start_map(
name="Wordcount with just mappers",
handler_spec="subset.mr.subset_mapper",
reader_spec="mapreduce.input_readers.DatastoreInputReader",
mapper_parameters={
"entity_kind": "data.subset."+kind,
"processing_rate": 500
},
mapreduce_parameters={model.MapreduceSpec.PARAM_DONE_CALLBACK:
'/subset/mr/callback'},
shard_count=shards,
queue_name="default",
)
record.mr_id = mapreduce_id
record.put()
self.redirect('/subset')