▸case-02 Our team needs to establish full system visibility for a new cloud-native platform before launching to production. Please prepare a comprehensive monitoring proposal that covers an evaluation of baseline capabilities, a target telemetry architecture design, a sequential rollout plan, defined metrics to capture, Grafana dashboard templates, operational alert runbooks, core SLO definitions, and developer service integration guides. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-08 We are instrumenting a distributed Node.js microservices architecture handling asynchronous Kafka events. Engineers are debating whether to pass trace contexts in custom HTTP headers like X-Trace-ID or rely on W3C Trace Context standard headers across HTTP and Kafka record headers. What is the recommended standard and context propagation strategy? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-13 We are deploying OpenTelemetry across 200 Kubernetes pods. An engineer proposes sending spans and metrics directly from application SDKs to an offsite SaaS backend over the public internet without an intermediate collector tier. What is the recommended OpenTelemetry collector deployment model for resilience and telemetry processing? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-21 Our PostgreSQL database cluster needs baseline observability. The database administrator suggests collecting only ping uptime checks. What critical metrics, slow query log configurations, and extensions should be configured for complete database visibility? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-05 Our PostgreSQL database CPU utilization spikes to 100% when running the order history search endpoint. Here is the SQL query and the EXPLAIN ANALYZE output showing a sequential scan on a 50-million row table. Please optimize this SQL query and suggest appropriate indexes. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-18 Our high-throughput payment platform generates 50TB of distributed trace data daily, creating unsustainable storage costs. The platform team wants to disable distributed tracing entirely. What trace sampling strategy balances cost reduction with retaining traces for system errors and high-latency events? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-19 Our SREs see a latency spike on a Grafana dashboard metric panel for http_request_duration_seconds_bucket, but they have to manually correlate the time window with Jaeger trace IDs to find the slow request. How can modern metric and trace integration solve this jump between metrics and traces? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-11 We want to alert on SLO consumption. A team member suggests triggering a P1 page whenever error rate spikes above 1% for 1 minute, but this creates constant false alarms during brief traffic bursts. What is the standard SRE approach for multi-window, multi-burn-rate alerting? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-14 Our Prometheus instance is crashing due to Out-Of-Memory (OOM) errors. We discovered that a developer added user_id and raw_email as labels on the http_requests_total metric. How should metric cardinality be managed in Prometheus monitoring? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-06 We need to set up an automated ArgoCD deployment manifest to execute progressive canary rollouts for our Kubernetes microservices. Please provide the Argo Rollouts CustomResourceDefinition (CRD) configuration for a 10% step canary strategy with automated rollback on health check failure. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-15 When setting up Prometheus Alertmanager alerts for high database replication lag, team members usually just put 'Check DB' in the alert summary. What essential sections must an operational alert runbook contain to reduce MTTR during an incident? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-09 Our checkout API targets 99.9% availability over a 30-day rolling window. The service receives 10,000,000 requests per month. A developer claims that 20,000 failed requests in a month is acceptable under this SLO. Is this correct, and what is the exact error budget in total allowed failed requests? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-22 We are storing 100GB of daily JSON logs in OpenSearch/Elasticsearch. The cluster disk is filling up because logs are kept indefinitely on hot nodes. How should an index lifecycle management (ILM) policy be structured for cost-effective log storage? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-20 We are writing SLOs for an asynchronous message processing worker pool reading from AWS SQS queues. The dev team proposes using CPU usage percentage as the primary Service Level Indicator (SLI). What is the appropriate SLI metric choice for queue-based worker processing? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-07 We are setting up baseline golden signals for an HTTP API gateway and a Redis caching tier. A junior engineer suggested measuring CPU percentage and disk space for the HTTP API gateway instead of request rates and latency. How should the metric catalog classify and differentiate resource metrics versus RED method metrics across these two tiers? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-16 We are launching a critical single-page e-commerce web application. The frontend team wants to rely solely on internal Prometheus node metrics to detect user-facing frontend page load delays and JavaScript errors. What frontend observability strategies should be implemented to capture real client performance? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-10 We are designing Grafana dashboards for a Kubernetes cluster running 50 microservices across 3 environments (dev, staging, prod). Should we build separate dashboard JSON files hardcoding each microservice name, or use dashboard variables and dynamic panels? Provide the design pattern. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-17 We are configuring Prometheus metrics collection for short-lived AWS Lambda functions that run for 500 milliseconds and terminate. A developer suggests configuring Prometheus to scrape the Lambda functions' HTTP endpoints directly every 15 seconds. Why does this fail, and what pattern handles ephemeral job metrics? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-04 Our payment service went down yesterday for 45 minutes due to a connection pool exhaustion on PostgreSQL. We have the raw database logs and thread dumps. Please conduct a root cause analysis post-mortem report detailing why the connection pool failed and how to patch the connection pool configuration. | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-12 Our Java microservices log unstructured plaintext to local disk files. We are deploying Elasticsearch and Fluentbit. Should we keep writing unstructured logs and parse them with complex regex at ingestion time, or change the application logger to JSON output with context MDC fields? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-01 We are migrating our legacy backend to microservices on Kubernetes and need a complete observability setup. Please deliver a blueprint that includes an assessment of monitoring requirements, a target architecture design, a step-by-step deployment roadmap, a metric dictionary, ready-to-use Grafana dashboard layouts, actionable alert response runbooks, service level targets with error budgets, and developer instrumentation instructions. | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
▸case-03 We are experiencing high mean time to resolution on our fintech payment APIs due to missing distributed tracing and metrics. Can you generate an observability guide covering an infrastructure evaluation, a comprehensive monitoring architecture, an implementation plan, critical metric definitions, dashboard templates, incident alert runbooks, target SLOs with error budget frameworks, and code-level integration instructions? | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |