A year of SRE work on a small Bitrix site (this post)
In this post, you’ll learn what I learned over a year when I picked up responsibility for my father’s website commercial and technical success. Despite all my Site Reliability experience, I wasted plenty of time and effort due to not following two simple rules I already knew, which are don’t change the system without looking at the relevant data and graphs you identified as a target of a change in advance, and, the primary rule of any refactoring, avoid introducing multiple changes at once as they make results interpolation close to impossible.
The website has a lot of decent content with SEO-friendly metadata, however, it did not have as good search engines positions as it deserved. One of the solid factors for search positions, especially for mobile users, is site loading speed. I thought it was the biggest problem I could fix to give the site an edge versus competitors.
I assumed that replacing Apache with Nginx + php-fpm, enabling Nginx PageSpeed mod and upgrading PHP version alongside checking the code for possible JS optimisations should speed the site up. The page readiness time graph below shows that part of the assumptions had the effect precisely opposite to the desired one.