Hello,
I have a question regarding replica shards. As I understand, a write to the primary gets propagated to replicas on other nodes via requests that use the same write pool and the same write queue as user-initiated writes on primaries on the node that contains the replica.
In case I increase the write queue size to a large number (for various reasons), is there any maximum time limit after which the primary would consider the replica as failed because its replication request is still waiting in the write queue while others are being processed on that node, or does it keep waiting until it gets a response?
(Elasticsearch version is 7.10 in case it is relevant)
Thank you.