We're running ES 2.4.x on NVMe cards for indexing and have disabled merge throttling:
curl -XPUT $esprodata01/_cluster/settings -d '{ "transient" : { "indices.store.throttle.type" : "none" }}'
{"acknowledged":true,"persistent":{},"transient":{"indices":{"store":{"throttle":{"type":"none"}}}}}
However the metric "merges.total_throttled_time_in_millis" goes up. Is there something I missed?