Performance
Measuring TTFB When a CDN Is Answering for You
Your TTFB chart says forty milliseconds and your origin says four hundred. Both are true. The CDN is answering the measurements, so you are benchmarking the…
DevOps
Requests, Limits and the QoS Class That Decided Who Died
When a node runs short of memory, Kubernetes does not choose eviction victims at random. It uses the QoS class, which your requests and limits silently…
Performance
The Index That Made Inserts Four Times Slower
Every index is a second write. On an insert heavy table, nine well meaning indexes turned a fast append into a small transaction against ten data…
Deep Dive
HTTP/2 Fixed Head of Line Blocking, Then Reintroduced It
HTTP/2 multiplexed many streams over one connection to escape head of line blocking, then moved the blocking one layer down into TCP, where a single lost…
Performance
Brotli Saved Forty Percent on Text and Nothing on Your Images
Switching to Brotli produced a celebratory chart and almost no user benefit, because the payloads that dominate the page were already compressed.
Tooling
The curl Flag That Tells You Which Part of the Request Was Slow
A slow request is five possible slow things. One curl flag with a write-out template splits the total into DNS, connect, TLS, time to first byte and…
Shopify
The Metaobject Query That Scanned Every Entry on Every Page
Metaobjects are great for structured merchant content, until a query with an unsupported filter loads every entry on every page view and filters them in Liquid.
Performance
PgBouncer Transaction Mode Broke Every Prepared Statement
After switching the pooler to transaction mode, the app started failing with 'prepared statement does not exist' at random.
DevOps
A Canary Deployment That Shipped a Broken Release Anyway
The canary looked healthy the entire time it was serving errors to real users. Five percent of traffic is not a sample, it is a rounding error, and your…
Tooling
perf record Turns a Mystery CPU Spike Into a Function Name
When CPU goes to 100 percent and your metrics show nothing, sampling the kernel's own profiler for thirty seconds will hand you the exact function responsible.
Shopify
The Section Rendering API Returned Yesterday's Cart Count
Section rendering is the right way to update a fragment without a full reload, and it is cacheable by default.
Performance
Your Average Latency Is Fine and Your Users Are Not
Mean latency is a number that describes nobody. Here is why the tail dominates real user experience, how retries multiply it, and which percentiles are…