Performance Panel wroclove.rb 2024

Panel discussion at wroclove.rb 2024 with Stephen Margheim, Maciej Rząsa, and Caio Almeida as panelists and an unnamed moderator. Covers typical Rails performance problems (database bottlenecks, action_view partials, loading too much data, serialization overhead in background jobs, joining data from multiple sources, growing models), how to triage issues (talk to business, find easy wins under 1-second thresholds, time-box explorations), tools (open telemetry, Honeycomb, CloudWatch, PG Analyzer, Apollo GraphQL, Grafana, Datadog, New Relic, OHA load testing, Speedscope, rbspy/stackprof profilers, Vernier, Prometheus+Grafana, Sidekiq dashboard, ActiveSupport Notifications, Sentry), heuristics for when to stop (define SLAs/performance budgets, time-box iteratively, use craftsmanship ad-hoc when no budget exists), war stories (Caio killed production by over-sanitizing parameters in a GraphQL migration; connection-pool contention from Rails-to-Python Lambda synchronous calls during Brazil 2022 elections project), and dealing with big filterable index endpoints (use Postgres with proper indexes, measure hot filter combinations, apply Pareto, avoid premature Elasticsearch adoption, negotiate UI limits). Date: 2024-03-22.