azure-bootstrap-primitives

Call configure_logging() from azure_bootstrap, which installs an ExtraFieldsFormatter and CorrelationFilter on the root logger and is idempotent. Use JsonLogFormatter for one JSON object per line when shipping to remote ingestion endpoints like Sumo Logic.

Wrap work in correlation_scope(correlation_id, **fields) from azure_bootstrap; the CorrelationFilter attaches every context var to each log record. It is built on contextvars, so it is safe for async tasks and concurrent requests.

DEBUG output requires a second gate: DEBUG_LOGGING_ENABLED must also be truthy, otherwise effective_log_level clamps the level to INFO. This defends against a stray manifest leaking DEBUG into production.

No. SumoLogicHandler appends to a bounded in-memory deque and a daemon thread performs the network I/O, so emit never blocks or raises. It retries 408/429/5xx with backoff and jitter, honors Retry-After, and drops oldest records on buffer overflow.

Transport factories return None as a soft no-op when the required extra or environment variable is missing, so enabling sumo_logic without the [sumologic] extra never errors. Install all transport dependencies with pip install 'azure-bootstrap[logging-all]'.

Use soft_fail or soft_fail_with to degrade a single operation to a fallback value while optionally re-raising UnrecoverableError. Use run_phases for a sequential pipeline that never aborts midway, returning PhaseResult objects with per-phase counters.