I was reading many articles online on how the replication protocol works in OpenSearch (Relevant to ES as most articles on that).
I was wondering if it’s possible in OpenSearch to define the minimum size for in-sync replicas. I’m aware of the
index.write.wait_for_active_shards feature, which will wait for say 2 / 3 active shards before continuing with writing to replicas, yet it’s not enforced - it those nodes failed the operation, it will still ack the write, although written say only to the primary.
My questions are:
- Is it possible to define a minimum number for the in-sync and if below per write, fail?
- What was the motivation for that decision in the protocol. It stands in contrast with other systems like Kafka, Cassandra, M3.
Would love any input on that,