Hi Folks,
We are seeing scenarios where the cluster is allocating 6 -10 shards(primaries & replicas) of an index with shard sizes around 50GB on a single node of the cluster even though all nodes in the cluster are of the same configuration. This behavior is causing issues when we have a high throughput of writes on that index coming in & the load is on that node which is housing these large shards on it & at times results in thread_pools maxing out & we encounter rejections
Want to understand what is the logic behind this behavior & is there any allocation algo that distributes the shards based on traffic ? so that the shard gets allocated to avoid the above mentioned scenario.