Reliability
Autonomous Remediation Without Compromising Production Safety
How Cordant defines the boundary between what an agent can safely do on its own and when it stops and asks a human.
The Runbook
Practical writing on incident response, SRE metrics, and production reliability. No product marketing. Just the operational problems that keep engineers up at 3 AM.
Incident Management
Zero Human Escalation: What It Actually Means for On-Call Teams
The goal is not to remove humans from the loop. It is to only wake them when their judgment is genuinely required.
Toil
Proactive vs Reactive: Rethinking SRE Toil from First Principles
Most toil reduction programs attack symptoms. The root cause is a response model that treats every alert as something a human must handle.
Observability
How Cordant Reads Production Signals Without Adding Alert Noise
Signal ingestion is only useful if it reduces the downstream decision burden. Here is how we approached the sampling and correlation problem.
Incident Response
The Case for Context-Aware Incident Triage
Identical alerts produce very different incidents depending on what changed in the last 4 hours. Most triage tools ignore that context entirely.
Tracing
What Distributed Tracing Misses in Real Production Incidents
Traces tell you what happened after a request entered your system. They are often silent about the infrastructure event that started the cascade.
Kubernetes
Kubernetes Health Probes Are Not a Substitute for Observability
Liveness and readiness probes are binary. Production degradation is not. Here is where the gap shows up.
On-Call
On-Call Load and Why We Keep Measuring It Wrong
Page count per shift is a lagging indicator of on-call health. By the time it moves, the burnout is already underway.
Platform Engineering
Alert Fatigue Is a System Design Problem, Not a Team Problem
Telling engineers to improve their signal-to-noise ratio assumes they control the alerting architecture. Most do not.
Failure Analysis
The Anatomy of a Cascading Failure
Cascades look chaotic in retrospect but almost always follow a readable sequence. Breaking down the dependency chain is where prevention starts.
Automation
Runbook Automation vs Incident Response: Two Different Problems
Automating a runbook gives you a faster human response. Automating incident response means the response happens before a human is involved.
SRE Metrics
Why MTTR Is the Wrong Metric to Optimize
Mean time to resolution hides the on-call experience inside a single number. Here is what to measure instead.
Putting the theory into practice
Cordant connects to your existing observability stack and handles the operational work that these articles describe. If you run production systems and want to see how it fits your setup, get in touch.