Build.com is a pure-play internet retailer, ranked # 2 in the home improvement market behind Home Depot. As an eCommerce site, Build.com has the need to address issues quickly and without interruptions to customers.
The CI/CD Initiative
Approximately three years ago Build.com kicked off their own CI/CD initiative. The result that followed was a mature Continuous Integration (CI) platform built using Jenkins. However, their CD platform was lacking automatic verification and rollback capabilities, which meant these tasks were often performed manually by team leads post-deployment. It was at this point that Ed Rose (Senior Director of Software Engineering at Build.com) and his team found and evaluated Harness.
Build.com needed a CD platform that could integrate with their Docker & Kubernetes environments, in addition to their existing ecosystem of tools that included Jenkins, Selenium, New Relic and Sumo Logic. “What we liked about Harness was that all our existing investment in CI/CD infrastructure was retained, and this allowed us to better utilize what we had already invested in.” said Ed.
Automatic Verification and Rollback
On day one of the evaluation, Ed’s team were able to migrate and reuse all their existing Jenkins investment inside of Harness and had a full CD pipeline built and running. For the auto-verification part of CD, Build.com required support for both New Relic and Sumo Logic. Two weeks later that support was delivered by Harness, which meant that Build.com was able to automate their entire CD process with automatic verification and rollback capabilities.
Build.com’s Mean-Time-To-Restore (MTTR) in production dropped from approximately 40 minutes (manual team-lead verification and custom rollback scripts) to approximately 10 minutes using Harnesses machine learning based auto-verification and rollback capabilities. Now 1 engineer at Build.com can verify production releases instantly instead of 6-7 team leads.
Canary Deployments Reduce Business Impact
Another Harness benefit Ed described was the ability for his team to automatically conduct canary-based deployments, thus significantly reducing the potential impact and risk of failed deployments over their previous blue/green deployments, which were being manually managed with higher risk and impact.
The Harness ROI
For Ed, the ROI was immediately clear, “We would need two full-time engineers with data science backgrounds for one year just to build the Harness verification capability, let alone the rest of Harness. We could easily spend $300,000 to do that, or just buy Harness and have a mature CD platform today” said Ed.
The end vision for Ed at Build.com is for his developers and teams to do their own deployments (CD As-A-Service) so they can release more frequently, and recover faster, should production deployments introduce regressions, anomalies or failure.
Watch the complete Harness interview with Ed Rose: