How about updating queries FROM the event store and not from the command handler ?

[jrouaix]

Hello guys

We were PoCing around your crate when found this :

cqrs/src/cqrs.rs

Line 154 in 09fc47b

for processor in &self.query_processors {

So the events are consumed to the projections from the command handler ?
What if the machine stops ?
What if the view database have issues ?
What will happen when the service restarts ?

[serverlesstechnology]

Hi @jrouaix,

You are correct that once any events are emitted from the command handler they become historical facts and are committed regardless of any query failures. This is essentially where the command phase of CQRS ends and the query phase begins. To keep the two phases decoupled, queries must manage failure conditions on their own.

For simple failure modes such as a view error you should deal with the error directly within the Query that you are using. If you are using a GenericQuery from the persist module this would entail noting the error (logs, telemetry, etc.) via an error handler in order to correct the issue.

To rectify more complex failures, a common solution is to use a field in a query to record the latest event that has been processed. Then, after a failure any query that is missing events can be identified and corrected.

A good mental model is that of these phases as two different services communicating via an event bus, and the same sort of failure recovery strategies apply.