I'm using Elasticsearch version 6.3 in production. We have a shard with size > 200GB in a data node. Recently, we did a "cluster reroute" wherein we moved this shard from node A (which had QueryCacheMemorySize = 3GB) to a new node B (which was in cluster but did not have any shards). After we did that the "QueryCacheMemorySize" increased to 9.2EB (the host obviously doesn't have this much memory) for the node B. We re-deployed, only to see the "QueryCacheMemorySize" go down to 0.
I am trying to see if there is some known issue, or a way to debug QueryCache related issues.