Some important points consider when developing an Apache Storm topology
- Kafka Consumer lag ---kafka tool to monitor
- Consumer group configuration of flux weaving YAML
- Storm UI metrics such as capacity < 1
- Spout parallelism should be equal to partition count of relevant topic
- Bolt parallelism should be increased according to the following equation
- **Throughput = Executor count * 1000/(process latency) * Capacity**
- Topology workers and supervisor server configurations accordingly
- Hbase database connection bottlenecks in Kerberized environment.
- Other Storm-related configuration parameters to be set in flux topology YAML or Ambari
- Maintain Tuple anchoring to guarantee the message processing through the downstream
References,
http://storm.apache.org/releases/1.0.6/Guaranteeing-message-processing.html
http://storm.apache.org/releases/1.0.6/Understanding-the-parallelism-of-a-Storm-topology.html
Comments
Post a Comment