when shard become very large, restart one cluster will waste too much time,
I know the primary shard become started is very fast.
but the cluster become green from yellow is very slow, become replica shard need recovery：
Especially the first phase of recovery.
Sometimes we need fast recovery services, not too concerned about data consistency。
I recommend adding a configuration item: “indices.recovery.fast”， which can be dynamically modified。
Then we can skip first phase of recovery~~
org.elasticsearch.indices.recovery.RecoverySourceHandler
public void phase1(final SnapshotIndexCommit snapshot) throws ElasticsearchException{
.....
//add one fast recovery branch
if (recoverySettings.fast() && !request.markAsRelocated()) {
......
} else if (recoverWithSyncId) {
......
} else {
......
}