aws/internal/log

Leveled logging for the SDK runtime.

Quiet by default, a firehose on demand (see PHILOSOPHY.md and the Logging section of RULES.md). The SDK says nothing on the happy path and only what an operator must see when something fails; turn the logger up to debug and it narrates both paths in full.

This is a thin pass-through to OTP logger (via aws_log_ffi) — the idiomatic BEAM mechanism, mirroring how the AWS Rust SDK leaves the sink to the application. There is no SDK-specific level knob: the operator configures verbosity and destination the standard way.

Levels — the logger’s primary level gates everything. It defaults to notice, under which error and warning are shown and debug is hidden. Turn on the firehose with, e.g., logger:set_primary_config(level, debug) or a kernel sys.config entry. (Per-module control via logger:set_module_level/2 also works — SDK events are emitted from the aws_log_ffi module.)
Destination — the logger’s handler decides. The default handler writes to standard_io (stdout); point it elsewhere (e.g. standard_error) via the handler config if you prefer. The SDK installs no handler of its own, so its lines pick up the host’s formatter and routing.

Values

debug

</>

pub fn debug(message: fn() -> String) -> Nil

Log at debug — the firehose, off unless the logger level is debug. The message is a thunk so its (often expensive) construction is skipped entirely when debug is not enabled.

error

</>

pub fn error(message: String) -> Nil

Log at error — unrecoverable, operator-must-see (credential chain exhausted, retries exhausted, the Lambda Runtime API gone fatal). Sparse, and on by default (outranks notice).

warning

</>

pub fn warning(message: String) -> Nil

Log at warning — notable but recovered (a retry fired, a credential provider was configured but failed). On by default (outranks notice).