def health_check(self):
# Save and load the current metrics.
endpoint_metrics, active_connections = self.update_metrics()
# Does a health check on all the endpoints that are being managed.
for endpoint in self.endpoints.values():
# Do not kick the endpoint if it is not currently owned by us.
if not(self.endpoint_owned(endpoint)):
continue
try:
metrics, metric_ips, endpoint_connections = \
self.load_metrics(endpoint, endpoint_metrics)
# Compute the active set (including custom metrics, etc.).
active = active_connections.get(endpoint.name, [])
active.extend(endpoint_connections)
active = list(set(active))
# Compute the globally weighted averages.
metrics = calculate_weighted_averages(metrics)
# Update the live metrics and connections.
logging.debug("Metrics for endpoint %s from %s: %s" % \
(endpoint.name, str(metric_ips), metrics))
self.zk_conn.write(paths.endpoint_live_metrics(endpoint.name), \
json.dumps(metrics), \
ephemeral=True)
self.zk_conn.write(paths.endpoint_live_active(endpoint.name), \
json.dumps(active), \
ephemeral=True)
# Run a health check on this endpoint.
endpoint.health_check(active)
# Do the endpoint update.
endpoint.update(reconfigure=False, metrics=metrics,
metric_instances=len(metric_ips),
active_ips=active)
except:
error = traceback.format_exc()
logging.error("Error updating endpoint %s: %s" % (endpoint.name, error))
try:
# Try updating our logs.
self.zk_conn.write(paths.manager_log(self.uuid), self.log.getvalue(), ephemeral=True)
except:
error = traceback.format_exc()
logging.error("Error saving logs: %s" % error)
# Reset the buffer.
self.log.truncate(0)
def get_manager_log(self, manager):
return self.zk_conn.read(paths.manager_log(manager))
