The DevOps Handbook – The Technical Practices of Feedback

Logging Levels

Understanding logging levels is crucial for effective system diagnostics and alerting. Alan Underwood and Joe Zack discuss the importance of using the correct logging levels, such as debug, info, warn, error, and fatal, to ensure meaningful data collection. They emphasize that misusing these levels can lead to unnecessary alerts, like being paged for low toner, which should be logged as info or warn rather than an error 1. Underwood highlights the significance of action-based logging for info levels, which capture system or user-initiated actions 2.

Debug logs are extremely verbose, capturing almost everything that happens in an application.

--- Joe Zack

Properly categorizing logs helps in maintaining system efficiency and avoiding alert fatigue 3.

   

Logging Challenges

Implementing effective logging systems in software applications presents several challenges. Alan Underwood and Joe Zack discuss the difficulty of integrating logging into code without cluttering it, especially in complex systems like Spark or Flink where extensive logging is necessary for debugging 4. They reference a Microsoft study showing that organizations with fewer server reboots had better service incident diagnosis, highlighting the importance of telemetry in both applications and deployment pipelines 5.

More metrics provide the confidence to change things.

--- Alan Underwood

This confidence is crucial for making informed changes and maintaining system reliability 6.