I'm making simple multi-search queries to elastic search 1.6, the queries are basic min max queries to sepecific daily indexes. When the indexes don't exist e.g. when searching for days in the future elastic search frequently hits the queue capacity 1000 query limit. But when within a normal index it does not. I can set threadpool.search.type to cached to fix this, but I'm worried about what is causing the difference between the two.
It's running on a 3 node cluster with 5 shards per index and min 2 node replication, where in my test I'm doing 8 queries in the multi-search.
I'm guessing this means it should be using 2 * 5 * 8 = 80 threads?
With further investigation I found the index name generator generated an empty string when the date exceeded now.
Meaning it quieried all indexes on the server, which I never want.
Looking at this: https://github.com/elastic/elasticsearch/issues/6470
I think it can't be disabled?