Hi, i'm logging time series events, and some clients read from this index sequentially, they take the top X items filtered by a fixed teanant query, and the next time they poll, the continue filtering by an orderer id > previous one.
Logstash act as bridge from rabbit, batching inserts.
The data is replicated for HA on 2 replicas.
I want to avoid eventual consistency for the latest inserted items, to avoid to skip them and skip forward next time.
Which is a good practice?
Can I filter always data inserted in the latest few finutes (1-2) to avoid missing data on replicas? I can monitor my nodes to undertand the speed rate of inserts, and mantain the 1s near real time buffer.
Otherwise can I route clients to read fixed replicas to let them read always data in the way is sync, then next time they will continue read without loose data?