def __init__(self, logs_dir, gcs_address):
"""Initialize the log monitor object."""
self.ip = services.get_node_ip_address()
self.logs_dir = logs_dir
self.publisher = gcs_pubsub.GcsPublisher(address=gcs_address)
self.log_filenames = set()
self.open_file_infos = []
self.closed_file_infos = []
self.can_open_more_files = True
def __init__(self, logs_dir, redis_address, redis_password=None):
"""Initialize the log monitor object."""
self.ip = services.get_node_ip_address()
self.logs_dir = logs_dir
self.redis_client = ray._private.services.create_redis_client(
redis_address, password=redis_password)
self.publisher = None
if gcs_pubsub.gcs_pubsub_enabled():
gcs_addr = gcs_utils.get_gcs_address_from_redis(self.redis_client)
self.publisher = gcs_pubsub.GcsPublisher(address=gcs_addr)
self.log_filenames = set()
self.open_file_infos = []
self.closed_file_infos = []
self.can_open_more_files = True
log_dir=args.logs_dir,
filename=args.logging_filename,
max_bytes=args.logging_rotate_bytes,
backup_count=args.logging_rotate_backup_count,
)
def is_proc_alive(pid):
try:
os.kill(pid, 0)
return True
except OSError:
# If OSError is raised, the process is not alive.
return False
log_monitor = LogMonitor(args.logs_dir,
gcs_pubsub.GcsPublisher(address=args.gcs_address),
is_proc_alive)
try:
log_monitor.run()
except Exception as e:
# Something went wrong, so push an error to all drivers.
gcs_publisher = GcsPublisher(address=args.gcs_address)
traceback_str = ray._private.utils.format_error_message(
traceback.format_exc())
message = (f"The log monitor on node {platform.node()} "
f"failed with the following error:\n{traceback_str}")
ray._private.utils.publish_error_to_driver(
ray_constants.LOG_MONITOR_DIED_ERROR,
message,
gcs_publisher=gcs_publisher,