t8y2/dbx · critical

Hive discovery returned no endpoints

Error message

Hive discovery returned no endpoints

What it means

Connect resolves HiveServer2 endpoints via service discovery (ZooKeeper). If the discovery process produced no recorded failures yet also yielded zero usable endpoints, the driver has nothing to connect to and returns this error. It distinguishes 'no endpoints at all' from 'all endpoints failed' (the latter produces the aggregated 'all HiveServer2 endpoints failed' error).

Source

Thrown at agents/drivers/argo-go/connector.go:165

					return connection, nil
				}
				rejected[target.address()] = true
				failures = append(failures, fmt.Sprintf("attempt %d %s: %v", attempt+1, target.address(), connectErr))
			}
			break
		}
		if attempt+1 < connector.retries && connector.retryInterval > 0 {
			timer := time.NewTimer(connector.retryInterval)
			select {
			case <-ctx.Done():
				timer.Stop()
				return nil, ctx.Err()
			case <-timer.C:
			}
		}
	}
	if len(failures) == 0 {
		return nil, errors.New("Hive discovery returned no endpoints")
	}
	return nil, fmt.Errorf("all HiveServer2 endpoints failed: %s", strings.Join(failures, "; "))
}

func (connector *discoveryConnector) Driver() driver.Driver {
	return connector.driver
}

func openHiveDatabase(config connectionConfig) *sql.DB {
	database := sql.OpenDB(newDiscoveryConnector(config))
	database.SetMaxOpenConns(1)
	database.SetMaxIdleConns(1)
	return database
}

func normalizeHiveAuth(value string) string {
	normalized := strings.ToUpper(strings.TrimSpace(value))
	switch normalized {

View on GitHub (pinned to c0390bff16)

Solutions

  1. Verify the ZooKeeper quorum hosts/paths in the connection config point to the namespace where HiveServer2 publishes (e.g. /hiveserver2)
  2. Confirm HiveServer2 is running with dynamic service discovery enabled and registered in ZooKeeper (zkCli.sh ls /hiveserver2)
  3. Retry once ZooKeeper is reachable — transient discovery failures can yield an empty endpoint set

Example fix

// before
dsn := "hive://user@zk1:2181,zk2:2181/default?serviceDiscovery=zooKeeper"
// after
dsn := "hive://user@zk1:2181,zk2:2181/default?serviceDiscovery=zooKeeper&zookeeperNamespace=/hiveserver2" // plus ensure HS2 is registered
Defensive patterns

Strategy: retry

Validate before calling

conn, _, err := zk.Connect(hosts, timeout)
if err != nil { return err }
children, _, err := conn.Children("/hiveserver2")
if err != nil { return err }
if len(children) == 0 { return errors.New("no HiveServer2 registered in ZooKeeper yet") }

Try / catch

endpoints, err := connector.Connect(ctx)
if err != nil {
	if strings.Contains(err.Error(), "no endpoints") {
		// back off and retry discovery; likely HS2 not yet registered
		time.Sleep(retryInterval)
		retry...
	}
}

Prevention

When it happens

Trigger: Calling Connect (or TestDiscoveryConnectorFailsOver/TestDiscoveryConnectorRetriesAllEndpoints flows) where the ZooKeeper service discovery node list is empty, all registration children are unparseable/skipped without error, or the endpoint iterator exits immediately with no attempts recorded.

Common situations: HiveServer2 is not publishing to ZooKeeper (zookeeper quorum misconfigured or hiveserver2.active_discovery not enabled); the znode path is wrong; dynamic service discovery mode not enabled on the Hive side; network partition leaves an empty/stale znode tree.

Related errors

AI-assisted analysis of t8y2/dbx@c0390bff16 (2026-09-05). Data as JSON: /api/errors/207f36b82ce60fb6. Report an issue: GitHub.