Closed cmwaters closed 1 year ago
A few questions came up for me while thinking about this issue
I have a skeleton implementation, but these questions stopped me up a bit from going further ! tia
Why are several of the methods no-op in the postgres implementation? I couldn't find this in the docs
The no-op methods are the searching methods. The reason they are no-op is because the underlying sink, e.g. postgres, already provides this functionality as that is the entire point of the sink. e.g. you wouldn't provide a SQL query directly to the RPC method, you'd use Postgres directly. The same would apply for Kafka or any other sink.
The implementation is quite simple, but I am running into this issue when writing the integrating tests.
https://github.com/segmentio/kafka-go/issues/684.
Will try sarama
package.
Summary
ADR 065 introduced custom TX indexing via event sinks. We also added postgres support which offers a rich query engine for clients (block providers / wallets) that want to make funky transaction queries. The other use case that clients may want transactional data is via event streaming.
Proposal
Apache Kafka is a commonly used tool that allows clients to subscribe to events. We should allow operators to hook up kakfa servers (producers) to a node via the
EventSink
. I'm not sure if we have to set up the topics or wether the operator has to but I imagine they just correlate to theEvent.Type
.For Admin Use