def __init__(self):
rpc_pb2_grpc.WorkerServicer.__init__(self)
etcdaddr = env.getenv("ETCD")
logger.info("using ETCD %s" % etcdaddr)
clustername = env.getenv("CLUSTER_NAME")
logger.info("using CLUSTER_NAME %s" % clustername)
# init etcdlib client
try:
self.etcdclient = etcdlib.Client(etcdaddr, prefix=clustername)
except Exception:
logger.error(
"connect etcd failed, maybe etcd address not correct...")
sys.exit(1)
else:
logger.info("etcd connected")
# get master ip and report port
[success, masterip] = self.etcdclient.getkey("service/master")
if not success:
logger.error("Fail to get master ip address.")
sys.exit(1)
else:
self.master_ip = masterip
logger.info("Get master ip address: %s" % (self.master_ip))
self.master_port = env.getenv('BATCH_MASTER_PORT')
self.imgmgr = imagemgr.ImageMgr()
self.fspath = env.getenv('FS_PREFIX')
self.confpath = env.getenv('DOCKLET_CONF')
self.taskmsgs = []
self.msgslock = threading.Lock()
self.report_interval = 2
self.lock = threading.Lock()
self.mount_lock = threading.Lock()
self.cons_gateway = env.getenv('BATCH_GATEWAY')
self.cons_ips = env.getenv('BATCH_NET')
logger.info("Batch gateway ip address %s" % self.cons_gateway)
logger.info("Batch ip pools %s" % self.cons_ips)
self.cidr = 32 - int(self.cons_ips.split('/')[1])
self.ipbase = ip_to_int(self.cons_ips.split('/')[0])
self.free_ips = []
for i in range(2, (1 << self.cidr) - 1):
self.free_ips.append(i)
logger.info("Free ip addresses pool %s" % str(self.free_ips))
self.gpu_lock = threading.Lock()
self.gpu_status = {}
gpus = gputools.get_gpu_status()
for gpu in gpus:
self.gpu_status[gpu['id']] = ""
self.start_report()
logger.info('TaskController init success')
def __init__(self):
rpc_pb2_grpc.WorkerServicer.__init__(self)
etcdaddr = env.getenv("ETCD")
logger.info ("using ETCD %s" % etcdaddr )
clustername = env.getenv("CLUSTER_NAME")
logger.info ("using CLUSTER_NAME %s" % clustername )
# init etcdlib client
try:
self.etcdclient = etcdlib.Client(etcdaddr, prefix = clustername)
except Exception:
logger.error ("connect etcd failed, maybe etcd address not correct...")
sys.exit(1)
else:
logger.info("etcd connected")
# get master ip and report port
[success,masterip] = self.etcdclient.getkey("service/master")
if not success:
logger.error("Fail to get master ip address.")
sys.exit(1)
else:
self.master_ip = masterip
logger.info("Get master ip address: %s" % (self.master_ip))
self.master_port = env.getenv('BATCH_MASTER_PORT')
# get worker ip
self.worker_ip = getip(env.getenv('NETWORK_DEVICE'))
logger.info("Worker ip is :%s"%self.worker_ip)
self.imgmgr = imagemgr.ImageMgr()
self.fspath = env.getenv('FS_PREFIX')
self.confpath = env.getenv('DOCKLET_CONF')
self.rm_all_batch_containers()
self.taskmsgs = []
self.msgslock = threading.Lock()
self.report_interval = 2
self.lock = threading.Lock()
self.mount_lock = threading.Lock()
self.gpu_lock = threading.Lock()
self.gpu_status = {}
gpus = gputools.get_gpu_status()
for gpu in gpus:
self.gpu_status[gpu['id']] = ""
self.start_report()
logger.info('TaskWorker init success')
def __init__(self, addr, etcdclient):
self.addr = addr
self.etcd = etcdclient
self.libpath = env.getenv('DOCKLET_LIB')
self.confpath = env.getenv('DOCKLET_CONF')
self.fspath = env.getenv('FS_PREFIX')
# set jupyter running dir in container
self.rundir = "/home/jupyter"
# set root running dir in container
self.nodehome = "/root"
self.lxcpath = "/var/lib/lxc"
self.imgmgr = imagemgr.ImageMgr()
self.historymgr = History_Manager()
def __init__(self, nodemgr, networkmgr, etcdclient, addr, mode, distributedgw='False'):
self.mode = mode
self.distributedgw = distributedgw
self.nodemgr = nodemgr
self.imgmgr = imagemgr.ImageMgr()
self.networkmgr = networkmgr
self.addr = addr
self.etcd = etcdclient
self.defaultsize = env.getenv("CLUSTER_SIZE")
self.fspath = env.getenv("FS_PREFIX")
self.clusterid_locks = threading.Lock()
# check database
try:
Container.query.all()
PortMapping.query.all()
VCluster.query.all()
except:
# create database
db.create_all()
logger.info ("vcluster start on %s" % (self.addr))
if self.mode == 'new':
logger.info ("starting in new mode on %s" % (self.addr))
# check if all clusters data are deleted in httprest.py
clean = True
usersdir = self.fspath+"/global/users/"
vclusters = VCluster.query.all()
if len(vclusters) != 0:
clean = False
for user in os.listdir(usersdir):
if len(os.listdir(usersdir+user+"/hosts")) > 0:
clean = False
if not clean:
logger.error ("clusters files not clean, start failed")
sys.exit(1)
elif self.mode == "recovery":
logger.info ("starting in recovery mode on %s" % (self.addr))
self.recover_allclusters()
else:
logger.error ("not supported mode:%s" % self.mode)
sys.exit(1)
G_networkmgr.printpools()
G_cloudmgr = cloudmgr.CloudMgr()
# start NodeMgr and NodeMgr will wait for all nodes to start ...
G_nodemgr = nodemgr.NodeMgr(G_networkmgr,
etcdclient,
addr=ipaddr,
mode=mode)
logger.info("nodemgr started")
distributedgw = env.getenv("DISTRIBUTED_GATEWAY")
G_vclustermgr = vclustermgr.VclusterMgr(G_nodemgr, G_networkmgr,
etcdclient, ipaddr, mode,
distributedgw)
logger.info("vclustermgr started")
G_imagemgr = imagemgr.ImageMgr()
logger.info("imagemgr started")
G_releasemgr = releasemgr.ReleaseMgr(G_vclustermgr, G_ulockmgr)
G_releasemgr.start()
logger.info("releasemgr started")
logger.info("startting to listen on: ")
masterip = env.getenv('MASTER_IP')
logger.info("using MASTER_IP %s", masterip)
masterport = env.getenv('MASTER_PORT')
logger.info("using MASTER_PORT %d", int(masterport))
G_historymgr = History_Manager()
master_collector = monitor.Master_Collector(G_nodemgr,