February 2018 Summaries

Integrating Harness with Dynatrace enhances automated verification of application performance across deployment pipelines, significantly reducing manual verification time and enabling automatic rollbacks in case of performance regressions. This integration allows customers to verify application performance automatically in various environments such as development, QA, staging, and production. By using continuous verification, companies like Build.com have drastically reduced the time spent on manual verifications from 18 hours a week to just 15 minutes per deployment, with Dynatrace providing real-time performance data analysis. Harness employs unsupervised machine learning models to detect performance deviations, and in cases of identified regressions, the system triggers automatic rollbacks to maintain application stability. The integration was facilitated using the Dynatrace API, allowing users to easily add Dynatrace verification to deployment workflows, with future enhancements expected to expand functionality further.

Feature flags, while offering flexibility and control in software development, can introduce cognitive load and technical debt if not managed properly. To mitigate these issues, it is crucial to implement clear expiration policies, automate backlogs, and limit the number of active flags. Retiring feature flags completely from both the configuration and the codebase ensures long-term success and reduces complexity. The concept of "zombie flags"—those that outstay their welcome—highlights the need for intentional management and established processes for flag retirement, such as setting expiration dates and enforcing their removal. Techniques like attaching time bombs to feature flags or applying Work in Progress (WIP) limits can help maintain a manageable number of active flags. The importance of removing flag implementations from the codebase is emphasized to prevent hidden costs. Effective feature flag management, as discussed, aims to keep development environments healthy and agile, allowing teams to leverage the benefits of feature flags without incurring unnecessary burdens.