Hello. I have two InfluxDB 1.6 instances. These use a lot of memory - usually around 60GB or so, each, but sometimes increasing to about 90-100GB. (This is heap bytes).
I’d like to understand where this is going. I can use SHOW STATS to account for about 20GB of memory, between TSM indexes (current shards are TSI but old ones are TSM) and the cache memBytes. That leaves a lot that I can’t account for. There are queries and writes, but there isn’t any correlation between running queries and high memory usage, and I’ve restricted it to 10 http writes at a time and the write batches are less than 10K points each, so I don’t think that would explain tens of GB of memory.
Can anyone give me some pointers for where to look or what config params I might tweak to try to reduce this?