Hi all, Running Cassandra 3.9 on 5 node cluster (c4.2xlarge machines).
We have loaded quite huge 350M rows table. Also there are some other read and write operations happening *on other small tables*, but the load is very little, cpu is normally less than 10%, network traffic from and to nodes is less than 10 Mb/sec for each node. Now when I run read intensive query (like select count(1) on that huge table) obviously Casandra is under pressure, cpu is high etc. If that's important the query is done via Presto. The query returns successfully after 40 minutes, but the cluster doesn't return to normal state. Cpu stays very high (~ 90%) for a very long period of time, and more strangely network traffic between nodes is more than 100 Mb/sec. Again, this is the situation after query has completed hours ago. Metrics doesn't show nothing suspicious like read repairs or compactions. Please advice what should I look at to understand the root cause of that high network traffic and cpu usage. Thanks in advance, Andrii