python-observability

Configure structlog with a processor chain including TimeStamper, add_log_level, and JSONRenderer, then call structlog.get_logger() to emit JSON logs. Add structlog.contextvars.merge_contextvars to automatically include correlation IDs in every entry.

Define Counter, Histogram, and Gauge metrics from prometheus_client for request count, latency, and resource usage, then increment them in a decorator or middleware around each endpoint. Use bounded label values like method, endpoint, and status.

Store the ID in a ContextVar, set it in middleware from the incoming X-Correlation-ID header or a new UUID, and bind it via structlog.contextvars. Pass the header explicitly on outbound httpx requests to downstream services.

Unbounded label values like user IDs create a new time series per value, exploding storage and query costs. Use bounded labels such as endpoint or user tier, and put high-cardinality identifiers in logs instead.

Yes, OpenTelemetry supports async Python through tracer.start_as_current_span context managers that work in async functions. Configure a TracerProvider with BatchSpanProcessor and an OTLP exporter to send spans to your backend.