jonsequitur / alluvial Goto Github PK

this makes it hard to differentiate different distributor logs from one another.

This mechanism should be fairly customizable, e.g. on a per-stream or per-aggregator basis, we should be able to specify when we can move on and when we should retry, etc. The Pipeline and Catch methods may be a good foundation for implementing this, though the API should be more expressive.

The streamId parameter of the method is not used, resulting in the incorrect stream id being passed to aggregators. The fix is streamId just needs to be passed to the Map call, So change from
.Map(es => es.Select(e => e.ToDomainEvent()));
to
.Map(es => es.Select(e => e.ToDomainEvent()), streamId);

It will be useful to measure performance, for example queries per second, throughput, and time to completion for projections.

The various Trace methods provide hooks that could be used for this.

Queries should be able to be distributed such that two different nodes (whether running on one machine or, ideally, distributed across machines) perform non-overlapping queries when catching up.

Each node will presumably need its own cursor.