There are two threads to send heartbeat in the Monitor and DiagnosisAgent.

[workingloong]

The elastic training agent only need to start one thread to send heartbeat. We need to remove the duplicated codes.

dlrover/dlrover/python/elastic_agent/diagnosis/diagnosis_agent.py

Lines 240 to 252 in c40dcee

	def send_heartbeat(self):
	try:
	ts = int(time.time())
	action = self._client.report_heart_beat(ts)
	self._agent_context.enqueue_diagnosis_action(action)
	except Exception as e:
	logger.warning(f"fail to report a heartbeat: {e}")
	def _periodically_report(self):
	logger.info("Start diagnosis agent reporter.")
	while True:
	self.send_heartbeat()
	time.sleep(15)

dlrover/dlrover/python/elastic_agent/monitor/training.py

Lines 126 to 141 in c40dcee

	def send_heartbeat(self):
	try:
	ts = int(time.time())
	action = self._master_client.report_heart_beat(ts)
	if action:
	pass
	except Exception:
	logger.warning("Fail to report a heartbeat.")
	def _periodically_report(self):
	logger.info("Start training agent reporter.")
	while True:
	if self._group_rank == 0:
	self.report_resource_with_step()
	self.send_heartbeat()
	time.sleep(15)