Fault-tolerance features

Beside struggling to provide a simple and lean API one of the main points of Rembus is its ability to be fault-tolerant respect to networks and application failures.

The Macro-based API and the component method provide an automatic reconnection policy in case of network faults and try at the best to guarantee message delivery when faults happen.

This mean that the following RPC service will run forever and it will reconnect automatically in case of network failures or broker unavailability.

using Rembus

@component "mycomponent"

function myservice(input::DataFrame)
    # run your super-cool logic and get back the result
    output_df = my_logic(input)
    return output_df
end

@expose myservice

@forever

Fault-tolerance holds equally for publish/subscribe setups: in case of connection lost the subscriber retries to reconnect to the broker until the connection will be up again.

If the subscription use the before_now option then messages published whereas the component was offline are delivered ordered by time of publishing when the component get online again.

using Rembus

@component "consumer"

function mytopic(input::DataFrame)
    # consume the dataframe posted to mytopic topic
end

@subscribe mytopic before_now

@forever