5.2.0 — 2026-09-07¶

Added¶

OtelMetricsExporter enum - Unique metrics exporter via METRICS_EXPORTER (otlp | pull) with optional pull bind host/port.
- otlp (default): push via periodic OTLP export
- pull: dedicated scrape HTTP server on METRICS_PULL_HOST:METRICS_PULL_PORT/metrics
OtelLogsExporter enum - Unique logs exporter via LOGS_EXPORTER (console | otlp); default console splits INFO/DEBUG → stdout and WARNING+ → stderr
LOGS_ENABLED default true - Log export on when OTel is enabled (console by default; switch to otlp for collectors)
METRICS_PULL_PORT default 8200 - Default scrape bind port for pull metrics
SYSTEM_METRICS_ENABLED - Kill switch for SystemMetricsInstrumentor (default true when metrics are on)

Metrics pull scrape lifecycle - Owned scrape server start/stop tied to OtelUtils init, shutdown, fork reset, and test reset
gRPC RPC duration metrics - Sync/async rpc.server.duration server interceptors installed by AppUtils when metrics are enabled (independent of traces)
metric_status_for_exception - Aligns decorator/RPC metric status with span status (BaseError HTTP < 500 → ok)
Duration histogram quality - Description plus explicit second bucket boundaries on function and gRPC duration histograms

opentelemetry-exporter-prometheus - Added to the otel extra to implement the pull scrape exporter

Pull scrape BDD - Scrape 200, pull-only, invalid exporter enum, disabled port, idempotent init, shutdown
Console logs BDD - Invalid LOGS_EXPORTER, INFO → stdout, WARNING → stderr
Metrics gaps BDD - gRPC rpc.server.duration without hand-rolled decorators, metrics-only gRPC, client BaseError metric status, system metrics off

Push vs pull - Documented METRICS_EXPORTER, scrape security/prefork notes, gRPC RPC metrics, status alignment, and instrument-name lowercasing gotchas