Async API: Fail-fast or Future-transported exceptions?

[mp911de]

The synchronous and reactive API are consistent in handling exceptions:

• sync calls throw instances of RedisExceptions on invocations
• calls to Observable<T> transport exceptions within on onError

The asynchronous API sometimes throws exceptions on invoking a Redis command and sometimes exceptions are carried within the RedisFuture (CompletionStage).

The question here is, whether to shift all exceptions to the future or keep it as-is.

Examples for exception cases:

• Redis connection is disconnected, and auto-reconnect is disabled: future-transported RedisException: Connection is in a disconnected state and reconnects disabled. Commands are not accepted.
• Connection is closed (by user): command invocation throws RedisException: Connection is closed
• Invalid command args (nulls, serialization issues): command invocation throws IllegalArgumentException
• Request queue size exceeded: command invocation throws RedisException: Request queue size exceeded
• Exceptions while request processing (Exceptions in the transport, IOExceptions, SSL Handshake errors, ... [except disconnections]): future-transported exceptions
• Exceptions while response processing (Deserialization, Exceptions in the transport): future-transported exceptions

[mp911de]

Changed "Redis connection is disconnected, and auto-reconnect is disabled" exception to fail-fast since that exception came in in 3.3/4.0 and both versions are not released now.

[mp911de]

Closing this ticket. The strategy is fail-fast over Future-transported exceptions.