Primary node of replica set showing 30% CPU usage when idle (on Linux)

We have a new installation of Mongo 6.0.14, on 3 different nodes (docker-based on each node). On the secondaries, we observe a very low load when idle (1%) but on the primary we see 300% (meaning 3 out of 4 cores) being used. When using “top -H -p ” we see that this load is produced by three threads named connXXXXX (X being numbers), which, through the logs, we are able to track down to the secondaries and the docker interface itself (I cannot explain that, yet). Is this something expected? In windows we have not seen such behavior.

This can be closed: The problem seems to be with the connections comming from metricbeat with replstatus.