High ops/sec but no active_requests

We’ve been using Couchbase 4.6 Community edition for a while now and are very satisfied so thank you for this great product. We have a two node cluster setup with two buckets, currently one bucket with around 17 million documents and the other with around 680k documents. We currently get peaks of high CPU usage, and one bucket has a high Ops/sec value (sometimes reaching 20k) but while querying ‘system:active_requests’ we don’t get any active requests, even though the ops/sec value is high.
Is it possible to get an explanation of the ops/sec value and why there is such a high cpu usage even though no queries are being made, are there any background operations we should be aware of?
Some technical details: Each node has around 32GB of ram to a total of 62.7 GB Ram, with about 437 GB free disk space on both nodes. Ram usage is around 50% for both nodes.

I believe system:active_requests only covers N1QL queries, not direct Key/Value operations.