Skip to content

Latest commit

 

History

History
52 lines (46 loc) · 4.14 KB

errors.md

File metadata and controls

52 lines (46 loc) · 4.14 KB

Error reference

Every exception derives from fanferro.FanferroError and has a retryable attribute (whether retrying the same operation unchanged may succeed). Errors raised by the Rust core also have category; AWS errors have service, operation, kind, code, aws_message, request_id and status.

Invalid options of the client, resources, consumers, policies and middleware raise ConfigurationError where they are given (a misspelt codec name lists the registered ones), and settling a message twice raises LeaseError. Both also derive from the built-in exception that Python code expects for them, ValueError and RuntimeError, as HandlerTimeoutError derives from TimeoutError. Arguments of the wrong type raise TypeError.

All of them are classes of fanferro.errors (also exported by fanferro), the module that tracebacks, the error_type of events and dead-letter metadata, and OpenTelemetry's error.type name: fanferro.errors.AccessDeniedError.

Errors pickle with their attributes, so they reach the caller of a worker process (ProcessPoolExecutor, multiprocessing) or a task queue as themselves: a ProvisioningError with its .report, a BatchPublishError with its .result. A value that does not pickle is left out (None in the unpickled error), and only that value: a failed message whose body is an instance of a class defined in a function arrives with body=None but with its id, type and attributes, a failure's error that does not pickle as error=None. Copies (copy.copy, copy.deepcopy) keep every attribute. AttributeValue and WireMessage pickle too.

Exception Raised when Retryable
ConfigurationError Invalid configuration: client, consumer, resource declarations, policies, codec names (also a ValueError) no
ValidationError A value is invalid before any network call (attributes, FIFO group id, reserved names) no
MessageTooLargeError Encoded message exceeds the destination limit without offloading (.size, .limit) no
AwsError An AWS call failed (kind: throttling, transient, timeout, network, not_found, access_denied, invalid_request, kms, receipt_handle_invalid, service) by kind
ResourceNotFoundError Queue/topic/subscription/object does not exist no
AccessDeniedError Missing permissions or credentials no
ThrottledError Throttled after SDK retries; also SQS refusing to create a queue again within 60 seconds of deleting it (QueueDeletedRecently) or to purge it again within 60 seconds (PurgeQueueInProgress) yes
TransportError Network/HTTP/timeout/checksum failure outside the SDK yes
PayloadError Storing or fetching a claim-check payload failed (also: URL not allowed) no
PayloadIntegrityError Downloaded payload size or SHA-256 does not match no
EnvelopeError Envelope could not be produced or parsed no
LeaseError Message already settled, or its lease is no longer valid (also a RuntimeError) no
ShutdownError Consumer/client closed, or used after fork() no
ProvisioningError Provisioning failed with failure mode "raise" (.report) by cause
ResourceUnavailableError A resource could not be resolved (log/ignore mode); re-resolved later yes
EncodeError / DecodeError (CodecError) A codec could not encode/decode no
RoutingError Invalid handler signature or route configuration no
BatchPublishError Some batch entries failed after retries (.result) if all failures are
HandlerTimeoutError A handler exceeded handler_timeout (also a TimeoutError) yes
ConsumerStateError Invalid consumer state transition (start twice, no routes) no
BatchItemFailedError A batch handler listed the message in its BatchResult without an error of its own yes
BatchItemFailuresError Records of an SQS event failed in a fanferro.aws_lambda handler created with report_batch_item_failures=False (.item_identifiers); Lambda returns the batch to the queue yes

Handler control flow uses fanferro.Retry, Requeue, Reject and Discard (subclasses of HandlerAction, not FanferroError).