Struct RetriableMessage 

pub struct RetriableMessage<TKey, TMessage>
where
    TKey: JobKey,
    TMessage: Message,
{
    pub key: TKey,
    pub message: Option<TMessage>,
    pub strategy: MessageRetryStrategy,
    pub retry_hook: Option<Arc<dyn Fn(&TKey) + Send + Sync + RefUnwindSafe>>,
    /* private fields */
}

A retriable message is a job message which will automatically be resubmitted to the factory in the event of a factory worker dropping the message due to failure (panic or unhandled error). This wraps the inner message in a struct which captures the drop, and if there’s still some retries left, will reschedule the work to the factory with captured state information.

IMPORTANT CAVEATS:

1. Regular loadshed events will cause this message to be retried if there is still retries left and the job isn’t expired unless you explicitely call completed() in the discard handler.
2. Consumable types are not well supported here without some wrapping in Option types, which is because the value is handled everywhere as &mut ref due to the drop implementation requiring that it be so. This means that RPCs using crate::concurrency::oneshots likely won’t work without some real painful ergonomics.
3. Upon successful handling of the job, you need to mark it as completed() at the end of your handling or discarding logic to state that it shouldn’t be retried on drop.

Fields

key: TKey

The key to the retriable job

message: Option<TMessage>

The message, which will be retried until it’s completed.

strategy: MessageRetryStrategy

The retry strategy

retry_hook: Option<Arc<dyn Fn(&TKey) + Send + Sync + RefUnwindSafe>>

A function which will be executed upon the job’s retry flow being executed (helpful for logging, etc)

SAFETY: We utilize std::panic::catch_unwind here to best-effort prevent a potential user-level panic from causing a process abort. However if the handler logic panics, then the retry hook won’t be executed, but the job will still be retried.

Implementations

impl<TKey, TMessage> RetriableMessage<TKey, TMessage>

where TKey: JobKey, TMessage: Message,

pub fn new( key: TKey, message: TMessage, strategy: MessageRetryStrategy, ) -> RetriableMessage<TKey, TMessage>

Construct a new retriable message with the provided number of retries. If the retries are None, that means retry forever. [Some(0)] will mean no retries

pub fn set_retry_hook( &mut self, f: impl Fn(&TKey) + Send + Sync + RefUnwindSafe + 'static, )

Attach a handler which will be executed when the job is retried (resubmitted to the factory).

pub fn from_job( _: Job<TKey, TMessage>, strategy: MessageRetryStrategy, factory: ActorRef<FactoryMessage<TKey, RetriableMessage<TKey, TMessage>>>, ) -> Job<TKey, RetriableMessage<TKey, TMessage>>

Convert a regular Job into a RetriableMessage capturing all the necessary state in order to perform retries on drop.

• job: The Job to convert
• strategy: The MessageRetryStrategy to use for this retriable message
• factory: The ActorRef of the factory which the job will be submitted to.

Returns a Job with message payload of RetriableMessage which can be retried should it be dropped prior to having job.msg.completed() called.

pub fn capture_retry_state( &mut self, options: &JobOptions, factory: ActorRef<FactoryMessage<TKey, RetriableMessage<TKey, TMessage>>>, )

Capture the necessary state information in order to perform retries automatically

• options: The JobOptions of the original job
• factory: The ActorRef of the factory the job will be submitted to

pub fn completed(&mut self)

Mark this message to not be retried upon being dropped, since it was handled successfully.

IMPORTANT: This should be called at the end of the handling logic, prior to returning from the worker’s message handler AND/OR from the discard logic. If you don’t provide a custom discard handler for jobs, then they will be auto-retried until a potential TTL is hit (or forever) and you may risk a spin-lock where an TTL expired job is submitted back to the factory, expire shedded again, dropped, and therefore retried forever. This is especially dangerous if you state that jobs should be retried forever.

Trait Implementations

impl<TKey, TMsg> Debug for RetriableMessage<TKey, TMsg>

where TKey: JobKey, TMsg: Message,

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl<TKey, TMessage> Drop for RetriableMessage<TKey, TMessage>

where TKey: JobKey, TMessage: Message,

fn drop(&mut self)

Executes the destructor for this type.