Check_MK Agent causing high Disk IO

Even after changing the APT plugin to async (every 900 sec) the IO is still at 1.5MB/s according to PVE.

Here is a graph from one PVE node. The writes per VMs are totally off.

The last 2 VMs are only that low because we just enabled cmk monitoring yesterday around 22h.

Currently I am out of ideas. :-/