The Site Reliability Podcast with Fexingo: SRE, Uptime, and Production Engineering

In episode 29 of The Site Reliability Podcast, Lucas and Luna dive into the specific metrics SRE teams use to reduce mean time to resolve (MTTR) during incidents. They break down the difference between mean time to acknowledge (MTTA) and MTTR, using real-world examples from companies like Google and Etsy. Lucas explains the concept of a 'rescue time' target—a hard limit on how long an incident can...

When a single service fails, the whole system shouldn't collapse. In this episode, Lucas and Luna dive into the circuit breaker pattern — a critical resilience tool in site reliability engineering. They break down how Netflix's Hystrix inspired modern implementations, how companies like Amazon and Lyft use circuit breakers to isolate failures, and why a poorly tuned breaker can make an outage wors...

In this episode of The Site Reliability Podcast, Lucas and Luna dive into the practical mechanics of error budgets — the SRE tool that lets teams trade reliability for feature velocity without breaking trust. They walk through a real example: a team running a service with a 99.9% SLO that has 0.1% error budget per month, and what happens when they burn through it by week two. Lucas explains how Go...

In Episode 26 of The Site Reliability Podcast, Lucas and Luna explore how SRE teams run 'game days' — simulated incident exercises — to build muscle memory and reduce panic during real outages. They break down how Etsy, a pioneer in game days, structures its exercises using realistic scenarios, mini-game design, and post-mortem debriefs without blame. The hosts discuss the difference between chaos...

In this episode of The Site Reliability Podcast, Lucas and Luna explore how SRE teams use error budgets to make smart trade-offs between reliability and feature velocity. They break down the concept with concrete examples from Google's original SRE model, showing how a 99.99% uptime target translates to 52.6 minutes of allowed downtime per year. The hosts discuss how error budgets empower teams to...

Most SRE teams have runbooks. Few have runbooks that engineers actually use in the middle of an incident. Lucas and Luna dive into why the typical runbook fails — too long, too vague, or written for the person who already knows the system. They break down what Google's internal SRE teams do differently: five-sentence maximum per procedure, explicit decision trees, and a 'runbook owners' workflow t...

Mean time to acknowledge (MTTA) is the clock that starts when an alert fires and stops when an engineer clicks 'ack'. For most teams, that gap is the single biggest waste of incident response time. In this episode, Lucas and Luna examine how Airbnb's SRE team cut their MTTA from 12 minutes to under 90 seconds by redesigning alert routing and escalation policies. They walk through the three-tier sy...

Episode 22 of The Site Reliability Podcast explores synthetic monitoring — proactive testing that catches outages before real users feel them. Lucas and Luna break down how companies like Etsy and Twilio simulate user journeys from multiple locations every minute, generating tens of thousands of transactions daily to validate critical flows. They discuss the difference between synthetic and real-u...

In this episode of The Site Reliability Podcast, Lucas and Luna explore traffic shadowing: a technique that lets SRE teams test new services with live production traffic without affecting real users. They break down how GitHub used shadowing to validate a new caching layer without risking customer data, and how Stripe employs it to test payment processing changes safely. Lucas explains the differe...

In this episode of The Site Reliability Podcast, Lucas and Luna dive into the practice of canary deployments—a key strategy for reducing blast radius in production. They break down how teams like Etsy and Netflix use phased rollouts to catch issues early, with specific numbers: Etsy's Deployinator halved deployment failures after adopting canaries, and Netflix's Spinnaker pipeline automatically ro...

Most incident response is reactive—you get paged, you triage, you fix. But a growing number of SRE teams are flipping the model: using historical data, machine learning, and anomaly detection to predict incidents before they actually impact users. In this episode, Lucas and Luna explore how companies like Google, Datadog, and a major European bank are deploying predictive SRE. They break down the...

In this episode, Lucas and Luna explore how site reliability engineering teams use capacity planning to avoid catastrophic outages during peak traffic events like Black Friday and Cyber Monday. They break down the specific methodology used by major e-commerce platforms, including the concept of 'headroom targets' and 'traffic shaping' — techniques that go beyond simple auto-scaling. Lucas explains...

Lucas and Luna explore how service level objectives (SLOs) have evolved from a technical metric into a strategic business tool. Using examples from Google, Etsy, and a mid-size fintech startup, they show how SLOs help SRE teams align with product managers, trade reliability for feature velocity, and communicate risk in terms executives understand. The episode drills into the concept of 'SLO-based...

Episode 16 of The Site Reliability Podcast explores toil budgets: the SRE practice of capping manual, repetitive work so teams have time for automation. Lucas and Luna break down how Google defined toil in its SRE book, how a mid-size fintech used a 50% toil budget to reduce incident response time, and why tracking toil by hand feels ironic. They discuss a concrete case where one team freed up 30...

Episode 15 of The Site Reliability Podcast with Fexingo dives into the human side of site reliability engineering: on-call burnout. Lucas and Luna explore how teams at companies like Etsy and Honeycomb use structured rotations, incident-free shifts, and proactive 'time to recover' metrics to keep engineers fresh. They break down specific data—like the effect of 12-hour versus 7-day rotations on al...

Lucas and Luna explore how site reliability engineers adapt chaos engineering beyond Netflix's famous Simian Army. The episode focuses on a mid-size e-commerce company, BlinkMart, which used controlled failure injection to uncover a critical database replication bug that would have caused a 45-minute outage during Black Friday. Lucas explains the difference between literal chaos—randomly killing s...

Episode 13 of The Site Reliability Podcast explores how Microsoft's SRE teams automate capacity planning to keep Azure running smoothly despite unpredictable demand. Lucas and Luna break down the three-layer approach — demand forecasting, headroom management, and autoscaling — and walk through a real case where a retail giant's Black Friday traffic spike was absorbed without a single incident. The...

Episode 12 of The Site Reliability Podcast with Fexingo digs into GitHub's postmortem culture — specifically how their SRE team runs incident reviews that actually prevent recurrence without destroying psychological safety. Lucas and Luna walk through the five-part structure GitHub uses, the distinction between 'blame' and 'accountability,' and why writing a timeline before identifying causes chan...

In this episode of The Site Reliability Podcast, Lucas and Luna dive into how Cloudflare's SRE team manages to process over 46 million HTTP requests per second across its global edge network. They explore the concept of 'edge of network' infrastructure, the role of anycast routing in distributing load, and how the team uses automated canary deployments to catch failures before they impact customer...

In Episode 10 of The Site Reliability Podcast, Lucas and Luna explore how AWS built a real-time observability platform that spots anomalies before they become customer-facing outages. They break down the three-tier architecture AWS uses to reduce mean time to detection, the role of distributed tracing at Amazon's scale, and why traditional monitoring falls short when you run millions of containers...

Netflix doesn't wait for failures to happen — it creates them on purpose. In this episode, Lucas and Luna explore how the streaming giant uses chaos engineering, specifically its tool Chaos Monkey, to build a system that withstands outages before they reach users. They break down the philosophy behind failure injection, how Netflix testing has evolved from simple instance termination to complex ne...

In this episode of The Site Reliability Podcast, Lucas and Luna dive into the specific engineering decisions Stripe made to reduce its mean time to remediate from roughly 25 minutes to under 15 minutes between 2023 and early 2026. They break down the three-layer incident response system Stripe built: a real-time chat-based escalation framework, an automated runbook generator called 'Cascade,' and...

In this episode, Lucas and Luna explore how Google's Site Reliability Engineering team uses psychological safety and structured incident management to reduce cognitive load and improve response times during outages. They break down the real story of a 2023 Google Cloud outage that lasted 47 minutes, how the incident commander role works in practice, and why 'blameless postmortems' are more than ju...

Lucas and Luna dig into a 2025 operational analysis from PagerDuty that put a dollar figure on alert fatigue: each unnecessary notification costs an engineering team roughly $1.08 per alert when you factor in context-switching time, cognitive load, and burnout-related attrition. They walk through how PagerDuty arrived at that number using data from over 6,000 incident response sessions, and why th...

SRE teams dread on-call rotation more than the pager itself. In this episode, Lucas and Luna explore a counterintuitive fix pioneered at Shopify: a fair-division algorithm based on game theory and the Talmud's approach to bankruptcy. No, really. The system, called 'Shapley-in-the-Cloud,' treats each engineer's preferred shifts as assets and uses the Shapley value to distribute the least desirable...