class Worker(plugins.SimplePlugin): def __init__(self, bus, port, options): plugins.SimplePlugin.__init__(self, bus) create_pycurl_thread(bus) if options.aux_port is not None: create_tcp_server(options.aux_port) self.id = None self.port = port self.master_url = options.master self.master_proxy = MasterProxy(self, bus, self.master_url) self.master_proxy.subscribe() if options.hostname is None: self.hostname = self.master_proxy.get_public_hostname() else: self.hostname = options.hostname self.lighty_conf_template = options.lighty_conf if options.blockstore is None: self.static_content_root = tempfile.mkdtemp( prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: self.static_content_root = options.blockstore block_store_dir = os.path.join(self.static_content_root, "data") try: os.mkdir(block_store_dir) except: pass self.block_store = BlockStore( self.hostname, self.port, block_store_dir, ignore_blocks=options.ignore_blocks) self.block_store.build_pin_set() self.block_store.check_local_blocks() create_watcher_thread(bus, self.block_store) self.upload_deferred_work = DeferredWorkPlugin(bus, 'upload_work') self.upload_deferred_work.subscribe() self.upload_manager = UploadManager(self.block_store, self.upload_deferred_work) self.execution_features = ExecutionFeatures() #self.task_executor = TaskExecutorPlugin(bus, self, self.master_proxy, self.execution_features, 1) #self.task_executor.subscribe() self.scheduling_classes = parse_scheduling_class_option( options.scheduling_classes, options.num_threads) self.multiworker = MultiWorker(ciel.engine, self) self.multiworker.subscribe() self.process_pool = ProcessPool( bus, self, self.execution_features.process_cacheing_executors) self.process_pool.subscribe() self.runnable_executors = self.execution_features.runnable_executors.keys( ) self.server_root = WorkerRoot(self) self.pinger = Pinger(bus, self.master_proxy, None, 30) self.pinger.subscribe() self.stopping = False self.event_log = [] self.log_lock = Lock() self.log_condition = Condition(self.log_lock) self.cherrypy_conf = {} cherrypy.config.update({"server.thread_pool": 20}) if options.staticbase is not None: self.cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } self.subscribe() def subscribe(self): self.bus.subscribe('stop', self.stop, priority=10) self.bus.subscribe("worker_event", self.add_log_entry) def unsubscribe(self): self.bus.unsubscribe('stop', self.stop) self.bus.unsubscribe("worker_event", self.add_log_entry) def netloc(self): return '%s:%d' % (self.hostname, self.port) def as_descriptor(self): return { 'netloc': self.netloc(), 'features': self.runnable_executors, 'has_blocks': not self.block_store.is_empty(), 'scheduling_classes': self.scheduling_classes } def set_master(self, master_details): self.master_url = master_details['master'] self.master_proxy.change_master(self.master_url) self.pinger.poke() def start_running(self): app = cherrypy.tree.mount(self.server_root, "", self.cherrypy_conf) if self.lighty_conf_template is not None: lighty = LighttpdAdapter(ciel.engine, self.lighty_conf_template, self.static_content_root, self.port) lighty.subscribe() # Zap CherryPy's original flavour server cherrypy.server.unsubscribe() server = cherrypy.process.servers.FlupFCGIServer( application=app, bindAddress=lighty.socket_path) adapter = cherrypy.process.servers.ServerAdapter( cherrypy.engine, httpserver=server, bind_addr=lighty.socket_path) # Insert a FastCGI server in its place adapter.subscribe() ciel.engine.start() if hasattr(ciel.engine, "signal_handler"): ciel.engine.signal_handler.subscribe() if hasattr(ciel.engine, "console_control_handler"): ciel.engine.console_control_handler.subscribe() ciel.engine.block() def stop(self): with self.log_lock: self.stopping = True self.log_condition.notify_all() def submit_task(self, task_descriptor): ciel.engine.publish("worker_event", "Start task " + repr(task_descriptor["task_id"])) ciel.engine.publish('execute_task', task_descriptor) def abort_task(self, task_id): ciel.engine.publish("worker_event", "Abort task " + repr(task_id)) self.task_executor.abort_task(task_id) def add_log_entry(self, log_string): with self.log_lock: self.event_log.append((datetime.now(), log_string)) self.log_condition.notify_all() def get_log_entries(self, start_index, end_index): with self.log_lock: return self.event_log[start_index:end_index] def await_log_entries_after(self, index): with self.log_lock: while len(self.event_log) <= int(index): if self.stopping == True: break self.log_condition.wait() if self.stopping: raise Exception("Worker stopping")
def master_main(options): deferred_worker = DeferredWorkPlugin(ciel.engine) deferred_worker.subscribe() worker_pool = WorkerPool(ciel.engine, deferred_worker, None) worker_pool.subscribe() task_failure_investigator = TaskFailureInvestigator(worker_pool, deferred_worker) job_pool = JobPool(ciel.engine, options.journaldir, None, task_failure_investigator, deferred_worker, worker_pool) job_pool.subscribe() worker_pool.job_pool = job_pool backup_sender = BackupSender(cherrypy.engine) backup_sender.subscribe() if options.hostname is not None: local_hostname = options.hostname else: local_hostname = socket.getfqdn() local_port = cherrypy.config.get('server.socket_port') master_netloc = '%s:%d' % (local_hostname, local_port) ciel.log('Local port is %d' % local_port, 'STARTUP', logging.INFO) if options.blockstore is None: static_content_root = tempfile.mkdtemp(prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: static_content_root = options.blockstore block_store_dir = os.path.join(static_content_root, "data") try: os.mkdir(block_store_dir) except: pass block_store = BlockStore(ciel.engine, local_hostname, local_port, block_store_dir) block_store.subscribe() block_store.build_pin_set() block_store.check_local_blocks() if options.master is not None: monitor = MasterRecoveryMonitor(cherrypy.engine, 'http://%s/' % master_netloc, options.master, job_pool) monitor.subscribe() else: monitor = None recovery_manager = RecoveryManager(ciel.engine, job_pool, block_store, deferred_worker) recovery_manager.subscribe() root = MasterRoot(worker_pool, block_store, job_pool, backup_sender, monitor) cherrypy.config.update({"server.thread_pool" : 50}) cherrypy_conf = dict() if options.staticbase is not None: cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } app = cherrypy.tree.mount(root, "", cherrypy_conf) lighty_conf_template = options.lighty_conf if lighty_conf_template is not None: lighty = LighttpdAdapter(ciel.engine, lighty_conf_template, static_content_root, local_port) lighty.subscribe() # Zap CherryPy's original flavour server cherrypy.server.unsubscribe() server = cherrypy.process.servers.FlupFCGIServer(application=app, bindAddress=lighty.socket_path) adapter = cherrypy.process.servers.ServerAdapter(cherrypy.engine, httpserver=server, bind_addr=lighty.socket_path) # Insert a FastCGI server in its place adapter.subscribe() if hasattr(ciel.engine, "signal_handler"): ciel.engine.signal_handler.subscribe() if hasattr(ciel.engine, "console_control_handler"): ciel.engine.console_control_handler.subscribe() ciel.engine.start() if options.workerlist is not None: master_details = {'netloc': master_netloc} master_details_as_json = simplejson.dumps(master_details) with (open(options.workerlist, "r")) as f: for worker_url in f.readlines(): try: post_string(urllib2.urlparse.urljoin(worker_url, 'control/master/'), master_details_as_json) # Worker will be created by a callback. except: ciel.log.error("Error adding worker: %s" % (worker_url, ), "WORKER", logging.WARNING) ciel.engine.block()
class Worker(plugins.SimplePlugin): def __init__(self, bus, hostname, port, options): plugins.SimplePlugin.__init__(self, bus) self.id = None self.hostname = hostname self.port = port self.master_url = options.master self.master_proxy = MasterProxy(self, bus, self.master_url) self.master_proxy.subscribe() if options.blockstore is None: block_store_dir = tempfile.mkdtemp(prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: block_store_dir = options.blockstore self.block_store = BlockStore(cherrypy.engine, self.hostname, self.port, block_store_dir, ignore_blocks=options.ignore_blocks) self.block_store.build_pin_set() self.upload_deferred_work = DeferredWorkPlugin(bus, 'upload_work') self.upload_deferred_work.subscribe() self.upload_manager = UploadManager(self.block_store, self.upload_deferred_work) self.execution_features = ExecutionFeatures() self.task_executor = TaskExecutorPlugin(bus, self.block_store, self.master_proxy, self.execution_features, 1) self.task_executor.subscribe() self.server_root = WorkerRoot(self) self.pinger = Pinger(bus, self.master_proxy, None, 30) self.pinger.subscribe() self.stopping = False self.event_log = [] self.log_lock = Lock() self.log_condition = Condition(self.log_lock) self.cherrypy_conf = {} if options.staticbase is not None: self.cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } if options.lib is not None: self.cherrypy_conf["/stdlib"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.lib } self.subscribe() def subscribe(self): self.bus.subscribe('stop', self.stop, priority=10) self.bus.subscribe("worker_event", self.add_log_entry) def unsubscribe(self): self.bus.unsubscribe('stop', self.stop) self.bus.unsubscribe("worker_event", self.add_log_entry) def netloc(self): return '%s:%d' % (self.hostname, self.port) def as_descriptor(self): return {'netloc': self.netloc(), 'features': self.execution_features.all_features(), 'has_blocks': not self.block_store.is_empty()} def set_master(self, master_details): self.master_url = master_details['master'] self.master_proxy.change_master(self.master_url) self.pinger.poke() def start_running(self): cherrypy.engine.start() cherrypy.tree.mount(self.server_root, "", self.cherrypy_conf) if hasattr(cherrypy.engine, "signal_handler"): cherrypy.engine.signal_handler.subscribe() if hasattr(cherrypy.engine, "console_control_handler"): cherrypy.engine.console_control_handler.subscribe() cherrypy.engine.block() def stop(self): with self.log_lock: self.stopping = True self.log_condition.notify_all() def submit_task(self, task_descriptor): cherrypy.engine.publish("worker_event", "Start task " + repr(task_descriptor["task_id"])) cherrypy.engine.publish('execute_task', task_descriptor) def abort_task(self, task_id): cherrypy.engine.publish("worker_event", "Abort task " + repr(task_id)) self.task_executor.abort_task(task_id) def notify_task_streams_done(self, task_id): self.task_executor.notify_streams_done(task_id) def add_log_entry(self, log_string): with self.log_lock: self.event_log.append((datetime.now(), log_string)) self.log_condition.notify_all() def get_log_entries(self, start_index, end_index): with self.log_lock: return self.event_log[start_index:end_index] def await_log_entries_after(self, index): with self.log_lock: while len(self.event_log) <= int(index): if self.stopping == True: break self.log_condition.wait() if self.stopping: raise Exception("Worker stopping")
def master_main(options): create_pycurl_thread(ciel.engine) deferred_worker = DeferredWorkPlugin(ciel.engine) deferred_worker.subscribe() worker_pool = WorkerPool(ciel.engine, deferred_worker, None) worker_pool.subscribe() task_failure_investigator = TaskFailureInvestigator(worker_pool, deferred_worker) job_pool = JobPool(ciel.engine, options.journaldir, None, task_failure_investigator, deferred_worker, worker_pool) job_pool.subscribe() worker_pool.job_pool = job_pool backup_sender = BackupSender(cherrypy.engine) backup_sender.subscribe() if options.hostname is not None: local_hostname = options.hostname else: local_hostname = socket.getfqdn() local_port = cherrypy.config.get('server.socket_port') master_netloc = '%s:%d' % (local_hostname, local_port) ciel.log('Local port is %d' % local_port, 'STARTUP', logging.INFO) if options.blockstore is None: static_content_root = tempfile.mkdtemp(prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: static_content_root = options.blockstore block_store_dir = os.path.join(static_content_root, "data") try: os.mkdir(block_store_dir) except: pass block_store = BlockStore(local_hostname, local_port, block_store_dir) block_store.build_pin_set() block_store.check_local_blocks() if options.master is not None: monitor = MasterRecoveryMonitor(cherrypy.engine, 'http://%s/' % master_netloc, options.master, job_pool) monitor.subscribe() else: monitor = None recovery_manager = RecoveryManager(ciel.engine, job_pool, block_store, deferred_worker) recovery_manager.subscribe() root = MasterRoot(worker_pool, block_store, job_pool, backup_sender, monitor) cherrypy.config.update({"server.thread_pool" : 50}) cherrypy_conf = dict() if options.staticbase is not None: cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } app = cherrypy.tree.mount(root, "", cherrypy_conf) lighty_conf_template = options.lighty_conf if lighty_conf_template is not None: lighty = LighttpdAdapter(ciel.engine, lighty_conf_template, static_content_root, local_port) lighty.subscribe() # Zap CherryPy's original flavour server cherrypy.server.unsubscribe() server = cherrypy.process.servers.FlupFCGIServer(application=app, bindAddress=lighty.socket_path) adapter = cherrypy.process.servers.ServerAdapter(cherrypy.engine, httpserver=server, bind_addr=lighty.socket_path) # Insert a FastCGI server in its place adapter.subscribe() if hasattr(ciel.engine, "signal_handler"): ciel.engine.signal_handler.subscribe() if hasattr(ciel.engine, "console_control_handler"): ciel.engine.console_control_handler.subscribe() ciel.engine.start() if options.workerlist is not None: master_details = {'netloc': master_netloc} master_details_as_json = simplejson.dumps(master_details) with (open(options.workerlist, "r")) as f: for worker_url in f.readlines(): try: post_string(urllib2.urlparse.urljoin(worker_url, 'control/master/'), master_details_as_json) # Worker will be created by a callback. except: ciel.log.error("Error adding worker: %s" % (worker_url, ), "WORKER", logging.WARNING) ciel.engine.block()
def master_main(options): deferred_worker = DeferredWorkPlugin(cherrypy.engine) deferred_worker.subscribe() global_name_directory = GlobalNameDirectory(cherrypy.engine) global_name_directory.subscribe() worker_pool = WorkerPool(cherrypy.engine, deferred_worker) worker_pool.subscribe() lazy_task_pool = LazyTaskPool(cherrypy.engine, worker_pool) task_pool_adapter = LazyTaskPoolAdapter(lazy_task_pool) lazy_task_pool.subscribe() task_failure_investigator = TaskFailureInvestigator(lazy_task_pool, worker_pool, deferred_worker) job_pool = JobPool(cherrypy.engine, lazy_task_pool, options.journaldir, global_name_directory) job_pool.subscribe() backup_sender = BackupSender(cherrypy.engine) backup_sender.subscribe() local_hostname = socket.getfqdn() local_port = cherrypy.config.get('server.socket_port') master_netloc = '%s:%d' % (local_hostname, local_port) print 'Local port is', local_port if options.blockstore is None: block_store_dir = tempfile.mkdtemp(prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: block_store_dir = options.blockstore block_store = BlockStore(cherrypy.engine, local_hostname, local_port, block_store_dir) block_store.build_pin_set() if options.master is not None: monitor = MasterRecoveryMonitor(cherrypy.engine, 'http://%s/' % master_netloc, options.master, job_pool) monitor.subscribe() else: monitor = None recovery_manager = RecoveryManager(cherrypy.engine, job_pool, lazy_task_pool, block_store, deferred_worker) recovery_manager.subscribe() scheduler = LazyScheduler(cherrypy.engine, lazy_task_pool, worker_pool) scheduler.subscribe() root = MasterRoot(task_pool_adapter, worker_pool, block_store, global_name_directory, job_pool, backup_sender, monitor, task_failure_investigator) cherrypy.config.update({"server.thread_pool" : 50}) cherrypy_conf = dict() if options.staticbase is not None: cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } cherrypy.tree.mount(root, "", cherrypy_conf) if hasattr(cherrypy.engine, "signal_handler"): cherrypy.engine.signal_handler.subscribe() if hasattr(cherrypy.engine, "console_control_handler"): cherrypy.engine.console_control_handler.subscribe() cherrypy.engine.start() if options.workerlist is not None: master_details = {'netloc': master_netloc} master_details_as_json = simplejson.dumps(master_details) with (open(options.workerlist, "r")) as f: for worker_url in f.readlines(): try: http = httplib2.Http() http.request(urllib2.urlparse.urljoin(worker_url, '/master/'), "POST", master_details_as_json) # Worker will be created by a callback. except: cherrypy.log.error("Error adding worker: %s" % (worker_url, ), "WORKER", logging.WARNING) cherrypy.engine.block()
class Worker(plugins.SimplePlugin): def __init__(self, bus, hostname, port, options): plugins.SimplePlugin.__init__(self, bus) self.id = None self.hostname = hostname self.port = port self.master_url = options.master self.master_proxy = MasterProxy(self, bus, self.master_url) self.master_proxy.subscribe() if options.blockstore is None: block_store_dir = tempfile.mkdtemp( prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: block_store_dir = options.blockstore self.block_store = BlockStore(cherrypy.engine, self.hostname, self.port, block_store_dir, ignore_blocks=options.ignore_blocks) self.block_store.build_pin_set() self.upload_deferred_work = DeferredWorkPlugin(bus, 'upload_work') self.upload_deferred_work.subscribe() self.upload_manager = UploadManager(self.block_store, self.upload_deferred_work) self.execution_features = ExecutionFeatures() self.task_executor = TaskExecutorPlugin(bus, self.block_store, self.master_proxy, self.execution_features, 1) self.task_executor.subscribe() self.server_root = WorkerRoot(self) self.pinger = Pinger(bus, self.master_proxy, None, 30) self.pinger.subscribe() self.stopping = False self.event_log = [] self.log_lock = Lock() self.log_condition = Condition(self.log_lock) self.cherrypy_conf = {} cherrypy.config.update({"server.thread_pool": 20}) if options.staticbase is not None: self.cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } if options.lib is not None: self.cherrypy_conf["/stdlib"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.lib } self.subscribe() def subscribe(self): self.bus.subscribe('stop', self.stop, priority=10) self.bus.subscribe("worker_event", self.add_log_entry) def unsubscribe(self): self.bus.unsubscribe('stop', self.stop) self.bus.unsubscribe("worker_event", self.add_log_entry) def netloc(self): return '%s:%d' % (self.hostname, self.port) def as_descriptor(self): return { 'netloc': self.netloc(), 'features': self.execution_features.all_features(), 'has_blocks': not self.block_store.is_empty() } def set_master(self, master_details): self.master_url = master_details['master'] self.master_proxy.change_master(self.master_url) self.pinger.poke() def start_running(self): cherrypy.engine.start() cherrypy.tree.mount(self.server_root, "", self.cherrypy_conf) if hasattr(cherrypy.engine, "signal_handler"): cherrypy.engine.signal_handler.subscribe() if hasattr(cherrypy.engine, "console_control_handler"): cherrypy.engine.console_control_handler.subscribe() cherrypy.engine.block() def stop(self): with self.log_lock: self.stopping = True self.log_condition.notify_all() def submit_task(self, task_descriptor): cherrypy.engine.publish( "worker_event", "Start task " + repr(task_descriptor["task_id"])) cherrypy.engine.publish('execute_task', task_descriptor) def abort_task(self, task_id): cherrypy.engine.publish("worker_event", "Abort task " + repr(task_id)) self.task_executor.abort_task(task_id) def notify_task_streams_done(self, task_id): self.task_executor.notify_streams_done(task_id) def add_log_entry(self, log_string): with self.log_lock: self.event_log.append((datetime.now(), log_string)) self.log_condition.notify_all() def get_log_entries(self, start_index, end_index): with self.log_lock: return self.event_log[start_index:end_index] def await_log_entries_after(self, index): with self.log_lock: while len(self.event_log) <= int(index): if self.stopping == True: break self.log_condition.wait() if self.stopping: raise Exception("Worker stopping")
class Worker(plugins.SimplePlugin): def __init__(self, bus, port, options): plugins.SimplePlugin.__init__(self, bus) self.id = None self.port = port self.master_url = options.master self.master_proxy = MasterProxy(self, bus, self.master_url) self.master_proxy.subscribe() if options.hostname is None: self.hostname = self.master_proxy.get_public_hostname() else: self.hostname = options.hostname self.lighty_conf_template = options.lighty_conf if options.blockstore is None: self.static_content_root = tempfile.mkdtemp(prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: self.static_content_root = options.blockstore block_store_dir = os.path.join(self.static_content_root, "data") try: os.mkdir(block_store_dir) except: pass self.block_store = BlockStore(ciel.engine, self.hostname, self.port, block_store_dir, ignore_blocks=options.ignore_blocks, aux_listen_port=options.aux_port) self.block_store.subscribe() self.block_store.build_pin_set() self.block_store.check_local_blocks() create_watcher_thread(bus, self.block_store) self.upload_deferred_work = DeferredWorkPlugin(bus, 'upload_work') self.upload_deferred_work.subscribe() self.upload_manager = UploadManager(self.block_store, self.upload_deferred_work) self.execution_features = ExecutionFeatures() #self.task_executor = TaskExecutorPlugin(bus, self, self.master_proxy, self.execution_features, 1) #self.task_executor.subscribe() self.scheduling_classes = parse_scheduling_class_option(options.scheduling_classes, options.num_threads) self.multiworker = MultiWorker(ciel.engine, self) self.multiworker.subscribe() self.process_pool = ProcessPool(bus, self) self.process_pool.subscribe() self.runnable_executors = self.execution_features.runnable_executors.keys() self.server_root = WorkerRoot(self) self.pinger = Pinger(bus, self.master_proxy, None, 30) self.pinger.subscribe() self.stopping = False self.event_log = [] self.log_lock = Lock() self.log_condition = Condition(self.log_lock) self.cherrypy_conf = {} cherrypy.config.update({"server.thread_pool" : 20}) if options.staticbase is not None: self.cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } self.subscribe() def subscribe(self): self.bus.subscribe('stop', self.stop, priority=10) self.bus.subscribe("worker_event", self.add_log_entry) def unsubscribe(self): self.bus.unsubscribe('stop', self.stop) self.bus.unsubscribe("worker_event", self.add_log_entry) def netloc(self): return '%s:%d' % (self.hostname, self.port) def as_descriptor(self): return {'netloc': self.netloc(), 'features': self.runnable_executors, 'has_blocks': not self.block_store.is_empty(), 'scheduling_classes': self.scheduling_classes} def set_master(self, master_details): self.master_url = master_details['master'] self.master_proxy.change_master(self.master_url) self.pinger.poke() def start_running(self): app = cherrypy.tree.mount(self.server_root, "", self.cherrypy_conf) if self.lighty_conf_template is not None: lighty = LighttpdAdapter(ciel.engine, self.lighty_conf_template, self.static_content_root, self.port) lighty.subscribe() # Zap CherryPy's original flavour server cherrypy.server.unsubscribe() server = cherrypy.process.servers.FlupFCGIServer(application=app, bindAddress=lighty.socket_path) adapter = cherrypy.process.servers.ServerAdapter(cherrypy.engine, httpserver=server, bind_addr=lighty.socket_path) # Insert a FastCGI server in its place adapter.subscribe() ciel.engine.start() if hasattr(ciel.engine, "signal_handler"): ciel.engine.signal_handler.subscribe() if hasattr(ciel.engine, "console_control_handler"): ciel.engine.console_control_handler.subscribe() ciel.engine.block() def stop(self): with self.log_lock: self.stopping = True self.log_condition.notify_all() def submit_task(self, task_descriptor): ciel.engine.publish("worker_event", "Start task " + repr(task_descriptor["task_id"])) ciel.engine.publish('execute_task', task_descriptor) def abort_task(self, task_id): ciel.engine.publish("worker_event", "Abort task " + repr(task_id)) self.task_executor.abort_task(task_id) def add_log_entry(self, log_string): with self.log_lock: self.event_log.append((datetime.now(), log_string)) self.log_condition.notify_all() def get_log_entries(self, start_index, end_index): with self.log_lock: return self.event_log[start_index:end_index] def await_log_entries_after(self, index): with self.log_lock: while len(self.event_log) <= int(index): if self.stopping == True: break self.log_condition.wait() if self.stopping: raise Exception("Worker stopping")
def master_main(options): deferred_worker = DeferredWorkPlugin(cherrypy.engine) deferred_worker.subscribe() global_name_directory = GlobalNameDirectory(cherrypy.engine) global_name_directory.subscribe() worker_pool = WorkerPool(cherrypy.engine, deferred_worker) worker_pool.subscribe() lazy_task_pool = LazyTaskPool(cherrypy.engine, worker_pool) task_pool_adapter = LazyTaskPoolAdapter(lazy_task_pool) lazy_task_pool.subscribe() task_failure_investigator = TaskFailureInvestigator( lazy_task_pool, worker_pool, deferred_worker) job_pool = JobPool(cherrypy.engine, lazy_task_pool, options.journaldir, global_name_directory) job_pool.subscribe() backup_sender = BackupSender(cherrypy.engine) backup_sender.subscribe() local_hostname = socket.getfqdn() local_port = cherrypy.config.get('server.socket_port') master_netloc = '%s:%d' % (local_hostname, local_port) print 'Local port is', local_port if options.blockstore is None: block_store_dir = tempfile.mkdtemp( prefix=os.getenv('TEMP', default='/tmp/sw-files-')) else: block_store_dir = options.blockstore block_store = BlockStore(cherrypy.engine, local_hostname, local_port, block_store_dir) block_store.build_pin_set() if options.master is not None: monitor = MasterRecoveryMonitor(cherrypy.engine, 'http://%s/' % master_netloc, options.master, job_pool) monitor.subscribe() else: monitor = None recovery_manager = RecoveryManager(cherrypy.engine, job_pool, lazy_task_pool, block_store, deferred_worker) recovery_manager.subscribe() scheduler = LazyScheduler(cherrypy.engine, lazy_task_pool, worker_pool) scheduler.subscribe() root = MasterRoot(task_pool_adapter, worker_pool, block_store, global_name_directory, job_pool, backup_sender, monitor, task_failure_investigator) cherrypy.config.update({"server.thread_pool": 50}) cherrypy_conf = dict() if options.staticbase is not None: cherrypy_conf["/skyweb"] = { "tools.staticdir.on": True, "tools.staticdir.dir": options.staticbase } cherrypy.tree.mount(root, "", cherrypy_conf) if hasattr(cherrypy.engine, "signal_handler"): cherrypy.engine.signal_handler.subscribe() if hasattr(cherrypy.engine, "console_control_handler"): cherrypy.engine.console_control_handler.subscribe() cherrypy.engine.start() if options.workerlist is not None: master_details = {'netloc': master_netloc} master_details_as_json = simplejson.dumps(master_details) with (open(options.workerlist, "r")) as f: for worker_url in f.readlines(): try: http = httplib2.Http() http.request( urllib2.urlparse.urljoin(worker_url, '/master/'), "POST", master_details_as_json) # Worker will be created by a callback. except: cherrypy.log.error( "Error adding worker: %s" % (worker_url, ), "WORKER", logging.WARNING) cherrypy.engine.block()