Hi! we upgraded to version 2.16 from version 2.7
we noticed a change in performance, slow queries and high cpu on the data nodes
are there any changes to do with all of those params maybe?
indices.query.bool.max_clause_count: 2400
#search_backpressure.mode: enforced
search_backpressure.mode: monitor_only
search_backpressure.node_duress.cpu_threshold: 0.95
search_backpressure.node_duress.heap_threshold: 0.95
search_backpressure.search_shard_task.total_heap_percent_threshold: 0.8
search_backpressure.search_shard_task.heap_percent_threshold: 0.8
search_backpressure.search_shard_task.cpu_time_millis_threshold: 600000
search_backpressure.search_shard_task.elapsed_time_millis_threshold: 600000
Could you tell me what happened to the cluster at 15:00 ~15:05? whether Bulk indexing or nothing.
Then the reason why CPU usage was very high in your data node would be the shortage of allocated CPU from VM(if you installed directly) or Worker Node(if the cluster is provisioned via k8s operator or helm chart).
How many vCores (and memory) did you allocate to your data node respectively?
When a large amount of data is being indexed, not only CircuitBreakingException occurred because of OOM, also CPU usage is almost full so can make other tasks of thread pool slow down.
thanks for the answer!
but expect for the spike that we had there a are having slow queries in general since the upgrade
Is there anything to check or do about it?