Handling Error Levels in Logging

In reviewing some provided code to a client, I observed a number of actions contray to generally accepted practices regarding logging. This is what I provided as the general programming conventions with regardings to logging.

Using Log4J (http://logging.apache.org/log4j/docs/ ), which is generally accepted as the benchmark for all Java applications. This provides the following logging levels.

  • FATAL
  • ERROR
  • WARN
  • INFO
  • DEBUG
  • TRACE – from 1.2.12, latest is 1.2.13

A description for what handling should occur per logging level.

FATAL. As the name suggests, all processing should stop. Should logging include a FATAL, the process is a Failure.

ERROR. An error has occured, and this requires attention, and action. Generally processing should stop, however additional post processing, or an alternative path could occur. Should logging include an ERROR, the process is a Failure.

WARN. Something that is unexpected occured, however it doesn’t affect the general processing from succeeding successfully. If a process includes WARN and not FATAL/ERROR it should be considered successful.

INFO. Information Only. On high volume systems, this level of logging may even be turned off. This generally indicates key information values or steps, and can assist when enabled in longer running processes to identify where a process is. You don’t log errors at INFO.

DEBUG. For Debugging Purposes only.

Ok, well that sounds like common sense. Here is what I observed on a client site.

  • Code logs a FATAL, but continues processing
  • A FATAL is logged, yet the calling process reports success
  • An ERROR is logged, yet the calling process reports success.
  • A lot of WARN are logged, and this is misleading, as it appears more information regarding XML elements not processed (We are talking 20+ Warnings per batch process). From what I’ve observed, these don’t require futher action, and should be changed in INFO.
  • Errors are being logged as INFO. A NullPointer RunTime Exception is logged as INFO. If an error provides an Exception argument where a stack trace is printed, it ain’t an INFO message.
Tagged with: General Java Open Source

Related Posts

Creating a More Realistic Benchmark

Common benchmark approaches fall into two general categories, synthetic testing and realistic testing. You have the most generic operations from a synthetic test, starting with the most simple example using a single table, a single column, and for a single DML operation.

Read more

Testing, Benchmarking, Evaluating

Testing and benchmarking are widely used terms in software technology, each serving a distinct purpose and goal. With the increasing adoption of AI in software development, the term evaluating has become significant and with this the re-emergence of what is quality assurance.

Read more

Your Attack Vector Extends Beyond Production Systems

A common data security issue is the unprotected copying of production data to non-production environments without any redaction, masking, or filtering. This practice poses a serious risk. A malicious actor will target the weakest link in your infrastructure, including non-production accounts and the developer systems accessing them.

Read more