tezvyn:

Monitoring & SRE

Observability, incident response, reliability, SLOs

263 bites

Monitoring & SRE2 min read

What is an error budget and how is it used?

WHAT IT TESTS: Understanding error budgets as a release governor. OUTLINE: The budget is the allowed unreliability (100 percent minus the SLO); track its burn, ship freely when budget remains, and freeze risky changes to focus on reliability when exhausted.

Monitoring & SRE2 min read

Defining SLIs and an SLO for an auth service?

WHAT IT TESTS: Translating user needs into measured reliability. OUTLINE: Pick user-centric SLIs like login availability and latency, measure good over valid events at the right boundary, then set an achievable SLO with a window.

Monitoring & SRE2 min read

Client-side chaos for an uncontrollable third party?

WHAT IT TESTS: Testing resilience you cannot reach directly. OUTLINE: Inject faults at your client boundary via a proxy or fault-injecting wrapper, simulate timeouts, errors, and latency, then verify timeouts, retries, breakers, and fallbacks.

Monitoring & SRE2 min read

Chaos test for gray-failure cascades in shared services?

WHAT IT TESTS: Detecting subtle cascading failure. OUTLINE: Inject partial latency into a shared service, hypothesize tenants stay isolated within steady state, and monitor cross-system queue depth, pool saturation, retries, and per-tenant SLIs.

Monitoring & SRE2 min read

Automating chaos in CI/CD for continuous verification?

WHAT IT TESTS: Operationalizing chaos as a gate. OUTLINE: Run codified experiments against staging or canary with pass/fail on steady-state SLIs; prerequisites are observability, automated abort, and isolation.

Monitoring & SRE2 min read

Resource faults versus network faults: when each matters?

WHAT IT TESTS: Matching fault type to a system's real risks. OUTLINE: Resource faults probe local saturation and autoscaling; network faults probe distributed-call resilience like timeouts and retries.

Monitoring & SRE2 min read

Why does 200ms latency drop requests? Diagnose it.

WHAT IT TESTS: Reasoning about concurrency limits and queueing. OUTLINE: Little's Law shows added latency raises in-flight requests, exhausting the thread or connection pool; check pool saturation, timeouts, and retries.

Monitoring & SRE2 min read

How do you run your first production chaos experiment?

WHAT IT TESTS: End-to-end experiment process and judgment. OUTLINE: Pick a low-risk known weakness, define a measurable hypothesis, brief stakeholders and on-call, run small with an abort, then analyze and fix.

Monitoring & SRE2 min read

Design a simple chaos experiment for a cache dependency?

WHAT IT TESTS: Applying chaos thinking to a cache. OUTLINE: Hypothesis that the service degrades gracefully when Redis is unavailable, monitor error rate, latency, DB load, and cache hit rate.

Monitoring & SRE2 min read

What is blast radius and how do you limit it?

WHAT IT TESTS: Safe production experimentation. OUTLINE: Blast radius is the scope of users or systems an experiment can harm; limit it by targeting a small traffic percentage and by having an automated abort.

Monitoring & SRE85 sec read

How does chaos engineering differ from other testing?

WHAT IT TESTS: Conceptual grasp of chaos engineering. OUTLINE: It experiments on real systems by injecting faults to test a steady-state hypothesis, versus verifying known behaviors like integration or load tests.

Monitoring & SRE2 min read

Reliability paved roads for an internal PaaS?

WHAT IT TESTS: Platform thinking that makes reliability the default. OUTLINE: Built-in observability, safe deploys with health checks and rollback, sane timeouts/retries/limits, and SLO tooling.

Monitoring & SRE2 min read

Design a chaos experiment for a payment dependency?

WHAT IT TESTS: Rigorous chaos experiment design. OUTLINE: Define a measurable steady state, hypothesize it holds when payments fail, limit blast radius to a small traffic slice, and auto-abort on SLO breach.

Monitoring & SRE88 sec read

Replication and consistency for active-active regions?

WHAT IT TESTS: Distributed consistency trade-offs. OUTLINE: Choose per data class between synchronous (low RPO, higher latency) and async replication, address write conflicts, and reason via CAP and PACELC.

Monitoring & SRE2 min read

Zero-downtime index migration on a hot table?

WHAT IT TESTS: Safe schema change at scale. OUTLINE: Build the index concurrently to avoid table locks, run off-peak with monitoring, and keep it reversible since dropping an index is cheap. RED FLAG: A blocking CREATE INDEX that locks writes on a hot table.

Monitoring & SRE86 sec read

Mitigating risk from an unproven external dependency?

WHAT IT TESTS: Designing for dependency failure. OUTLINE: Timeouts and circuit breakers to fail fast, bulkheads to isolate resources, fallbacks or cached/degraded responses.

Monitoring & SRE83 sec read

Reliability patterns for queue-based job processing?

WHAT IT TESTS: Designing fault-tolerant async workers. OUTLINE: Retries with backoff and jitter for transient faults, dead-letter queues plus a poison-message limit, idempotent handlers and visibility timeouts.

Monitoring & SRE82 sec read

What is an SLO and how do you define API availability?

WHAT IT TESTS: Understanding SLI versus SLO and how to measure reliability. OUTLINE: SLO is a target on an SLI, availability SLI is good requests over valid requests, you need labeled request counts over a window.

Monitoring & SRE88 sec read

What telemetry must a new microservice ship with?

WHAT IT TESTS: Practical observability literacy. OUTLINE: Metrics for trends like p99 latency and error rate, structured logs for per-request detail, traces for cross-service causality. RED FLAG: Listing only host CPU dashboards and no request-level signals.

Monitoring & SRE2 min read

When horizontal scaling is the wrong fix

WHAT IT TESTS: Scaling judgment. OUTLINE: adding instances fails when the bottleneck is a shared resource like one database, a lock, or a queue, so more instances just add contention; investigate where time is actually spent.