Best Practices for Java Logging and Custom Logback Appender

In Java, the common log levels are DEBUG, INFO, WARN, and ERROR; developers typically enable DEBUG in development and INFO in production, using popular frameworks such as Log4j or Logback.

The ultimate goal of logging is to enable rapid problem location and resolution, so the article presents several best‑practice recommendations for each level, including printing request/response data, method inputs and outputs, and contextual information for WARN and ERROR logs.

When applications run on multiple machines, locating logs becomes difficult. The article briefly describes the classic ELK architecture (Kafka → Logstash → Elasticsearch → Kibana) and stresses the importance of correlating logs across instances using a unified identifier such as a request ID.

To automatically add a request ID to every log entry, a custom Logback appender can be created. The steps are:

Extend AppenderBase and implement the start() method to initialize a Kafka producer.

Override the append() method, retrieve the original log message from the ILoggingEvent, prepend the request ID (usually stored in a ThreadLocal or TransmittableThreadLocal set by an HTTP filter), and forward the enriched message to Kafka.

This approach ensures that all logs contain a consistent request identifier, making it easy to filter and analyze logs for a specific transaction across a distributed environment.