def download_logs(self):
reservations = list()
ssh_username = self.config.globals.ssh_username
if self.reservations:
reservations = self.reservations
else:
for cloud in self.clouds:
reservations = cloud.conn.get_all_instances()
for reservation in reservations:
for instance in reservation.instances:
if self.database.check_benchmark(self.benchmark.name,
instance.id):
local_path = os.path.join(
self.config.globals.log_local_path,
self.benchmark.name, instance.id)
if not os.path.exists(local_path):
os.makedirs(local_path)
for path in self.path:
file_name = os.path.basename(path)
local_path = os.path.join(local_path, file_name)
now = (datetime.datetime.now()).strftime("%H%M%S")
local_path = local_path + '_' + now + '_' + \
instance.instance_type
com = "scp -r " + ssh_username + "@" + \
instance.public_dns_name + ":" + path + " " + \
local_path
LOG.debug("Download logs: [%s] download %s into %s" %
(self.benchmark.name, os.path.basename(path),
local_path))
command = Command(com)
command_return = command.execute()
if command_return != 0:
LOG.error("Download logs: " + command.stdout)
LOG.error("Download logs error: " + command.stderr)
def download_logs(self):
reservations = list()
ssh_username = self.config.globals.ssh_username
for cloud in self.clouds:
for instance in cloud.get_all_floating_ips():
if self.database.check_benchmark(self.benchmark.name,
instance.instance_id):
local_path = os.path.join(
self.config.globals.log_local_path,
self.benchmark.name, instance.instance_id)
if not os.path.exists(local_path):
os.makedirs(local_path)
for path in self.path:
file_name = os.path.basename(path)
local_path = os.path.join(local_path, file_name)
now = (datetime.datetime.now()).strftime("%H%M%S")
local_path = local_path + '_' + now + '_' + \
instance.instance_id
com = "scp -r " + ssh_username + "@" + \
instance.ip + ":" + path + " " + \
local_path
LOG.debug("Download logs: [%s] download %s into %s" %
(self.benchmark.name, os.path.basename(path),
local_path))
command = Command(com)
command_return = command.execute()
if command_return != 0:
LOG.error("Download logs: " + command.stdout)
LOG.error("Download logs error: " + command.stderr)
def _update_node_info(self):
self.nodes = []
self.num_total_nodes = 0
self.num_total_cores = 0
self.num_free_cores = 0
self.num_down_cores = 0
pbsnodes_cmd = str(self._pbsnodes_cmd) + " -a"
pbsnodes = Command([pbsnodes_cmd])
pbsnodes_rc = pbsnodes.execute()
if pbsnodes_rc != 0:
LOG.error("pbsnodes returned %d" % pbsnodes_rc)
return
node_line = "\n(\S+)\n\s+state\s=\s(\S+)\n\s+np\s=\s(\d+)\n"
node_pattern = re.compile(node_line)
matches = re.findall(node_pattern, pbsnodes.stdout)
for match in matches:
n = Node(match[0], int(match[2]), match[1])
self.num_total_nodes += 1
self.num_total_cores += int(match[2])
if match[1] == "free":
self.num_free_cores += int(match[2])
if "down" in match[1]:
self.num_down_cores += int(match[2])
if "down" not in match[1]:
if n.public_dns_name not in self._has_booted:
self._has_booted.append(n.public_dns_name)
self.nodes.append(n)
LOG.debug("Nodes updated: %s total nodes and %s total cores." % (
self.num_total_nodes, self.num_total_cores))
def download_logs(self):
reservations = list()
ssh_username = self.config.globals.ssh_username
if self.reservations:
reservations = self.reservations
else:
for cloud in self.clouds:
reservations = cloud.conn.get_all_instances()
for reservation in reservations:
for instance in reservation.instances:
if self.database.check_benchmark(self.benchmark.name, instance.id):
local_path = os.path.join(self.config.globals.log_local_path, self.benchmark.name, instance.id)
if not os.path.exists(local_path):
os.makedirs(local_path)
for path in self.path:
file_name = os.path.basename(path)
local_path = os.path.join(local_path,file_name)
local_path = local_path+'_'+(datetime.datetime.now()).strftime("%H%M%S")+'_'+instance.instance_type
com = "scp -r "+ssh_username+"@"+instance.public_dns_name+":"+path+" "+local_path
LOG.debug("Download logs: [%s] download %s into %s" % (self.benchmark.name, os.path.basename(path), local_path))
command = Command(com)
command_return = command.execute()
if command_return != 0:
LOG.error("Download logs: "+command.stdout)
LOG.error("Download logs error: "+command.stderr)
def _add_new_node(self, public_dns_name, np):
qmgr_cmd = str(self._qmgr_cmd) + " -c \"create node %s np=%d\""
qmgr_cmd = qmgr_cmd % (public_dns_name, np)
add_node = Command([qmgr_cmd])
add_node_rc = add_node.execute()
if add_node_rc != 0:
LOG.error("qmgr returned %d" % add_node_rc)
return
LOG.debug("Successfully added node: %s" % public_dns_name)
def scp_log_back(self):
scp_string = "scp %s@%s:~/%s %s/sleep.log" \
% (self.config.workload.user, self.master.dns, self.config.workload.log_remote, self.config.log_dir)
scp_cmd = Command(scp_string)
code = scp_cmd.execute()
if code == 0:
LOG.info("Successfully obtained the log from the master node")
else:
LOG.error("Error occurred during obtaining the log from the master node")
def _remove_node(self, public_dns_name):
qmgr_cmd = str(self._qmgr_cmd) + " -c \"delete node %s\""
qmgr_cmd = qmgr_cmd % public_dns_name
remove_node = Command([qmgr_cmd])
remove_node_rc = remove_node.execute()
if remove_node_rc != 0:
LOG.error("qmgr returned %d" % remove_node_rc)
return
if public_dns_name in self._has_booted:
self._has_booted.remove(public_dns_name)
LOG.debug("Successfully removed node: %s" % public_dns_name)
def offline_node(self, public_dns_name):
pbsnodes_cmd = str(self._pbsnodes_cmd) + " -o %s"
pbsnodes_cmd = pbsnodes_cmd % public_dns_name
offline_node = Command([pbsnodes_cmd])
offline_node_rc = offline_node.execute()
if offline_node_rc != 0:
LOG.error("pbsnodes returned %d" % offline_node_rc)
return
else:
LOG.debug("Successfully marked node offline: %s" % public_dns_name)
for node in self.nodes:
if node.public_dns_name == public_dns_name:
node.terminate_me = True
def obtain_end_time(self):
# cat /Users/dmdu/Dropbox/dmdu-Downscaling-Experiments/experiments/exprC/AP-Mix-Thresh-Infinite/20130224_215317/sleep.log | grep terminated | head -n 1
file_name = "%s/%s" % (self.directory, "sleep.log")
end_time_string = "cat %s | grep terminated | tail -n 1" % (file_name)
#print "Command: %s " % (end_time_string)
cmd = Command(end_time_string)
code = cmd.execute()
if code:
print "Can't obtain the end time"
print cmd.stderr
self.end = None
else:
line = cmd.stdout
items = line.split()
self.end = self.convert_time(items[3], False) #no time conversion; Chicago time
print "End time: %s" % (self.end)
def __init__(self, options):
#self.clouds = CloudsConfig(read_config(options.clouds_file))
#self.benchmarking = BenchmarkingConfig(read_config(options.benchmarking_file))
self.globals = GlobalConfig(options.global_file)
self.master = MasterConfig(options.master_file)
self.clouds = CloudsConfig(options.clouds_file)
self.phantom_config = PhantomConfig(options.phantom_file)
self.workers = WorkersConfig(options.workers_file)
self.workload = WorkloadConfig(options.workload_file)
self.policy = PolicyConfig(options.policy_file)
__timestamp = datetime.datetime.now()
timestamp = __timestamp.strftime("%Y%m%d_%H%M%S")
self.phantom_config.domain_name = "%s_%s" % (self.phantom_config.domain_name_prefix, timestamp)
self.experiment_id = timestamp
# Create directory for all the logs
self.log_dir = "log/%s" % (self.experiment_id)
os.mkdir(self.log_dir)
self.remote_log = "%s/%s" % (self.log_dir, options.remote_log)
self.node_log = "%s/%s" % (self.log_dir, options.node_log)
self.worker_pool_log = "%s/%s" % (self.log_dir, options.worker_pool_log)
self.discarded_work_log = "%s/%s" % (self.log_dir, options.discarded_work_log)
self.failure_log = "%s/%s" % (self.log_dir, options.failure_log)
# to keep current code running for now
self.threshold = self.policy.threshold
self.downscaler_interval = self.policy.downscaler_interval
# Copy config files to the log directory for current experiment
copy_string = "cp etc/* %s/" % (self.log_dir)
copy_cmd = Command(copy_string)
code = copy_cmd.execute()
if code == 0:
LOG.info("Config files have been copied successfully to the log directory")
def _update_job_info(self):
qstat = Command([self._qstat_cmd])
qstat_rc = qstat.execute()
if qstat_rc != 0:
LOG.error("qstat returned %d" % qstat_rc)
return
job_line = "(\S+)\s+(\S+)\s+(\S+)\s+(\S+)\s+(\S+)\s+"
job_line += "(\d+)\s+(\d+)\s+(\S+)\s+(\S+)\s+([A-Z])\s+(\S+)"
job_pattern = re.compile(job_line)
queued_cores = 0
queued_jobs = 0
total_jobs = 0
for line in qstat.stdout.split('\n'):
match = job_pattern.match(line)
if match:
if match.group(10) == 'Q':
queued_cores += int(match.group(7))
queued_jobs += 1
total_jobs += 1
self.num_queued_jobs = queued_jobs
self.num_queued_cores = queued_cores
self.num_total_jobs = total_jobs
LOG.debug("Jobs updated: %s total jobs and %s queued cores." % (
self.num_total_jobs, self.num_queued_cores))
def run(self):
batch_index = 0
for batch in self.batch_files:
last_line = os.popen("tail -n 1 %s" % batch).read()
# if sleep time is specified
if ("SLEEP" in last_line) or ("sleep" in last_line):
# last item in the line
sleep_time = int(last_line.split()[-1:][0])
else:
sleep_time = 0
# Copy the batch file to the log directory
copy_string = "cp %s %s/" % (batch, self.config.log_dir)
copy_cmd = Command(copy_string)
code = copy_cmd.execute()
if code == 0:
#LOG.info("Workload %s file has been copied successfully to the log directory" % (batch))
pass
# Scp this file to the master
scp_string = "scp %s %s@%s:~/%s" % (batch, self.config.workload.user, self.master.dns, self.config.workload.submit_remote)
scp_cmd = Command(scp_string)
code = scp_cmd.execute()
if code == 0:
#LOG.info("Batch file %s has been copied to the master node" % (batch))
pass
else:
LOG.error("Error occurred during copying batch file %s to the master node" % (batch))
# Send this batch to the work queue
exec_cmd = RemoteCommand(
config = self.config,
hostname = self.master.dns,
ssh_private_key = self.config.globals.priv_path,
user = self.config.workload.user,
command = 'condor_submit %s' % (self.config.workload.submit_remote))
code = exec_cmd.execute()
if code == 0:
#LOG.info("Batch file %s has been submitted to the work queue" % (batch))
pass
else:
LOG.error("Error occurred during submission of batch file %s" % (batch))
LOG.info("%s" % (batch))
batch_index += 1
# Sleep for a while if this is specified in the batch file
time.sleep(sleep_time)
# Periodic log saving, every 100 jobs
if batch_index == 100:
self.scp_log_back()
batch_index = 0
# To give it enough time so the jobs are scheduled; unless specified otherwise
if sleep_time == 0:
time.sleep(60)
# After this for loop, go into monitor mode (run while there are jobs in the queue)
LOG.info("Workload turns into monitor mode: this thread will stop when there are no more jobs in the queue. Sleep interval: %d" % (self.interval))
jobs = Jobs(self.config, self.master.dns)
count = jobs.get_current_number()
print "Initial job count: %d" % (count)
counter = 0
while count > 0:
time.sleep(self.interval)
count = jobs.get_current_number()
print "Job count: %d" % (count)
counter += 1
# Periodic log saving, every 20 iterations
if counter == 20:
self.scp_log_back()
counter = 0
# check that the queue is empty one more time
time.sleep(120)
LOG.info("Checking that the queue is empty one more time after 120 seconds")
count = jobs.get_current_number()
while count > 0:
time.sleep(self.interval)
count = jobs.get_current_number()
print "Job count: %d" % (count)
LOG.info("Workload completed")
