How to build efficient Kafka broker healthcheck?

Question

In my app I will perform some kind of health check of my Kafka cluster.

Currently I make a TopicMetadataRequest to detect dead brokers:

  Future {
    // this will fail if Kafka is unavailable
    consumer.send(new TopicMetadataRequest(Seq("health-check-topic"), 1))
  }

Unfortunately this call produces a huge network traffic, because of Cluster topology/settings.

Is there a better way to check kafka brokers? What I need is something simple like true/false indicator.

Answer 1

You can also use Zookeeper API to get the broker list as follows:

ZooKeeper zk = new ZooKeeper(KafkaContextLookupUtil.getZookeeperConnect().getZkConnect(), 10000, null);
    List<String> ids = zk.getChildren("/brokers/ids", false);
    List<Map> brokerList = new ArrayList<>();
    ObjectMapper objectMapper = new ObjectMapper();

    for (String id : ids) {
        Map map = objectMapper.readValue(zk.getData("/brokers/ids/" + id, false, null), Map.class);
        brokerList.add(map);
    }
    return brokerList;

Answer 2

If you want to build your own health check, this is a current (January 2020) list of KIPs covering health checks:

• KIP-143: Controller Health Metrics
• KIP-188: Add new metrics to support health checks
• KIP-237: More Controller Health Metrics

Regarding Harvinder Singh's currently accepted answer:

Kafka Manager is great but it's evolving slowly. There's of course Confluent Control Center - a part of Confluent Platform, but you'll need a license for it. Confluent is a company founded by the team that built Apache Kafka. I've heard about akHQ (ex KafkaHQ) (HackerNews story). Here's a list of management consoles maintained on Apache Kafka Confluence page (check URLs there):

• Kafka Manager - A tool for managing Apache Kafka.
• kafkat - Simplified command-line administration for Kafka brokers.
• Kafka Web Console - Displays information about your Kafka cluster including which nodes are up and what topics they host data for.
• Kafka Offset Monitor - Displays the state of all consumers and how far behind the head of the stream they are.
• Capillary - Displays the state and deltas of Kafka-based Apache Storm topologies. Supports Kafka >= 0.8. It also provides an API for fetching this information for monitoring purposes.
• Doctor Kafka - Service for cluster auto healing and workload balancing.
• Cruise Control - Fully automate the dynamic workload rebalance and self-healing of a Kafka cluster.
• Burrow - Monitoring companion that provides consumer lag checking as a service without the need for specifying thresholds.
• Chaperone - An audit system that monitors the completeness and latency of data stream.

If you don't need GUI, there are also:

• https://github.com/andreas-schroeder/kafka-health-check
• and its fork https://github.com/ustream/kafka-health-check