app

internal/app

Logic overview

The app package wires runtime role behavior (leader vs follower), scrape/backfill loops, shared health state, and HTTP endpoint composition.

• Runtime backend wiring selects shared dependencies from config:
  • Redis store when store.backend=redis and connection succeeds.
  • In-memory store fallback if Redis init fails.
  • RabbitMQ management API queue when AMQP config is present.
  • RabbitMQ startup topology declaration for exchange/main queue/DLQ/retry queues.
  • In-memory queue fallback if RabbitMQ init fails.
• Leader mode runs scheduled scrape cycles and writes metrics to the shared store.
• In multi-org mode, leader scheduling runs one ticker loop per org using each org's configured scrape_interval.
• Each leader cycle also writes gh_exporter_leader_cycle_last_run_unixtime so /metrics is non-empty even when no activity metrics are produced.
• Leader cycles enqueue targeted backfill for both whole-org failures and per-repo missed windows returned by scraper partial results.
• GitHub unhealthy cooldown gates scheduled scraping after repeated failures and enqueues github_unhealthy backfill jobs while cooling down.
• Leader cycles run store GC and emit operational metrics (scrape_runs_total, store_write_total, backfill enqueue counters, rate-limit metrics, dependency health).
• Follower mode consumes backfill jobs with configurable worker fan-out, re-scrapes missed windows, applies idempotency locks, and writes backfill processing metrics.
• Runtime continuously probes Redis and AMQP dependencies to drive readiness and health component state.
• Health status is evaluated from role-aware dependency flags and continuously exported as gh_exporter_dependency_health.
• HTTP routing exposes /metrics, /livez, /readyz, and /healthz, with /metrics backed by configurable snapshot caching (full or incremental refresh).
• Runtime logs include role startup context plus per-cycle scrape summaries for easier operational debugging.

API reference

Types

• Runtime: orchestrates store, queue, scraper manager, health status, and role loop lifecycle.
• RoleHooks: lifecycle contract used by role management (StartLeader, StopLeader, StartFollower, StopFollower).
• RoleManager: consumes leadership events and transitions between leader/follower hooks.

Functions

• NewRuntime(cfg *config.Config, orgScraper scrape.OrgScraper, logger ...*zap.Logger) *Runtime: constructs runtime, initializes configured store/queue backends with safe fallbacks, creates dispatcher, and wires scrape manager.
• NewHTTPHandler(metricsHandler http.Handler, healthHandler http.Handler) http.Handler: returns a chi router for metrics and health endpoints.
• NewRoleManager(hooks RoleHooks) *RoleManager: constructs a role manager.

Methods

• (*Runtime) Store() runtimeStore: returns the runtime metric store abstraction used by exporter/tests.
• (*Runtime) QueueDepth() int: returns queued backfill message depth.
• (*Runtime) Handler() http.Handler: returns combined metrics+health handler.
• (*Runtime) StartLeader(ctx context.Context): transitions runtime into leader loop.
• (*Runtime) StopLeader(): stops leader responsibilities.
• (*Runtime) StartFollower(ctx context.Context): transitions runtime into follower loop.
• (*Runtime) StopFollower(): stops follower responsibilities.
• (*Runtime) CurrentStatus(ctx context.Context) health.Status: returns evaluated health status.
• (*Runtime) RunLeaderCycle(ctx context.Context) error: performs one scrape pass, queues backfill for failures, and records internal cycle heartbeat metrics.
• (*RoleManager) Run(ctx context.Context, roleEvents <-chan bool, electorErrs <-chan error) error: processes role changes until context cancellation, stream close, or elector error.